Marco Dozza

ALMA MATER STUDIORUM
UNIVERSITÀ DI BOLOGNA
DEIS – Department of Electronics, Computer Science, and Systems
Biofeedback Systems for Human
Postural Control
A method for understanding sensory integration and improving motor training
Marco Dozza
Thesis submitted for the degree of Ph.D. in Bioengineering (ING-INF/06)
Supervisor
Prof. Angelo Cappello
Co-supervisors
Prof. Lorenzo Chiari
Prof. Fay B. Horak
Reviewer
Prof. Carlo Marchesi
(Relatore: Marco Dozza)
(Coordinatore: Prof. Angelo Cappello)
I
II
Keywords
Postural Control
Sensory Integration
Auditory-, Visual-, Tactile-Biofeedback
Accelerometry
Vestibular Loss
Motor Learning
III
IV
Thesis Abstract
Reliable sensory information and correct integration of sensory information are necessary
for the control of posture. When sensory information is inadequate, such as in vestibular
loss subjects, balance and postural control are impaired. Further, gradual loss of sensory
information is also a consequence of the natural process of aging and is one of the major
causes leading to falls in the elderly population.
Sensory information can be augmented by using biofeedback (BF) devices. BF devices
are artificial systems able to provide additional movement information to their users. Although
BF devices for postural control have been experimented since the 70s, the extent to which
BF devices can substitute for missing sensory information for the control of posture is still
unknown. Further, although BF devices have been suggested to be helpful for rehabilitation,
no study to date has provided conclusive evidence that practice with BF is better at improving
retention of motor performance than practice without BF.
The purpose of this dissertation are 1) to design, set-up and validate new-generation,
portable, low-cost BF devices, 2) to determine how the design features of such BF devices
influence postural control during static and dynamic motor tasks in vestibular loss and
healthy subjects, 3) to elucidate how movement information from BF devices is integrated
with sensory information for the control of posture, and 4) to understand the relationship
between the effect of BF and spontaneous motor learning on postural control.
We implemented several types of BF devices that coded postural sway from bi-axial
accelerometers, a combination of accelerometers and gyros, and a force plate into a stereo
sound, vibrotactile stimulation to the trunk and/or a visual representation using different
coding algorithms. By comparing such devices, we demonstrated how crucial the design
of a BF device is, since it influences both the motor performance and the postural strategy
of its user. In addition, we showed how vestibular loss and healthy subjects can use our
audio-BF device to reduce sway by augmenting postural control without increasing muscular
stiffness. Further, we found that audio-BF increases closed-loop control of posture and does
not influence the open loop control of posture.
By testing bilateral vestibular loss and healthy subjects in several conditions of limited or
inadequate sensory information, we showed how audio-BF efficacy is related to the individual
dependency of each subject on vestibular, somatosensory, and visual information. In addition,
we showed that audio-BF improves posture also in dynamic tasks such as standing on a
randomly rotating surface and that the extent of these postural improvements is proportional
V
to the amount of movement information coded into the sound. Also, we showed that
spontaneous motor learning and audio-BF affect different ranges of frequency of postural
control during standing on a randomly rotating surface.
Furthermore, unilateral vestibular loss subjects were tested during tandem gait using
a cross-over design to understand whether tactile-BF of trunk tilt could improve postural
performances during a complex, dynamic motor task such as gait. Results from this experiment
showed that tactile-BF of trunk tilt acts similarly to natural sensory feedback in immediately
improving dynamic motor performance and not as a method to recalibrate motor performance
to improve dynamic balance function after short-term use.
Our results have many implications for the design of BF devices, for the understanding
of motor control and sensory integration, and for the design of the protocols to be used
with BF devices. More specifically, our results suggest that BF 1) needs a customized design
for each subject and each task to optimally improve postural motor performance without
facilitating undesirable postural strategies, 2) improves motor control in static and dynamic
tasks by augmenting motor information and substituting for missing sensory information, 3)
must be equipped with training protocols able to favor motor learning in order to became a
helpful tool for balance and motor rehabilitation and training.
VI
VII
VIII
Index
I. Sensory Integration and Augmentation for the Control of Posture .................................................................................1
Abstract ..................................................................................................................................................................................................................................... 3
Evidence of Sensory Integration for the Control of Posture ..................................................................................................................................... 4
Concurrence and Interference of Sensory Information ............................................................................................................................................. 4
Tuning and Calibration of Sensory Information for Internal Model Representation .......................................................................................... 6
Movement Disorders and Their Relation to Sensorimotor Integration .................................................................................................................. 8
Sensory Loss and Aging .................................................................................................................................................................................................. 8
Augmentation of Sensory Information for Postural Control ................................................................................................................................... 11
Overview on Biofeedback Experimentation .............................................................................................................................................................. 11
Design of Biofeedback Systems ................................................................................................................................................................................... 12
Experimentation of Biofeedback Systems: Protocol Design ................................................................................................................................... 15
Bibliography .......................................................................................................................................................................................................................... 17
II. Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System ................................................... 25
Abstract ................................................................................................................................................................................................................................... 27
Introduction ........................................................................................................................................................................................................................... 28
Materials and Methods ....................................................................................................................................................................................................... 29
Sensory Unit..................................................................................................................................................................................................................... 29
Processing Unit................................................................................................................................................................................................................ 29
Audio Output Unit .......................................................................................................................................................................................................... 30
Algorithm for ABF Sound Generation ......................................................................................................................................................................... 30
Experimental Protocol and Results ................................................................................................................................................................................ 33
Discussion .............................................................................................................................................................................................................................. 36
Bibliography .......................................................................................................................................................................................................................... 38
III. Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway .......................... 41
Abstract ................................................................................................................................................................................................................................... 43
Introduction ......................................................................................................................................................................................................................... 44
Methods .................................................................................................................................................................................................................................. 46
Participants ...................................................................................................................................................................................................................... 46
Apparatus and procedure ............................................................................................................................................................................................. 46
Data recording ................................................................................................................................................................................................................ 47
Data analysis ................................................................................................................................................................................................................... 47
Statistical analysis .......................................................................................................................................................................................................... 48
Results .................................................................................................................................................................................................................................... 49
Subjects’ confidence and comfort ............................................................................................................................................................................... 49
Subjects’ sway .................................................................................................................................................................................................................. 49
IX
Table of Contents
Center of Pressure analysis ........................................................................................................................................................................................... 49
Acceleration analysis ..................................................................................................................................................................................................... 50
Stabilogram diffusion analysis .................................................................................................................................................................................... 51
Muscle activity analysis ................................................................................................................................................................................................. 51
Discussion ............................................................................................................................................................................................................................. 53
Bibliography .......................................................................................................................................................................................................................... 56
IV. Direction Specificity of Audio-Biofeedback for Postural Sway ...................................................................................... 59
Abstract ................................................................................................................................................................................................................................... 61
Introduction ........................................................................................................................................................................................................................... 62
Materials & Methods ........................................................................................................................................................................................................... 63
Results ..................................................................................................................................................................................................................................... 65
Discussion .............................................................................................................................................................................................................................. 67
Bibliography .......................................................................................................................................................................................................................... 69
V. Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss ...................................................... 71
Abstract ................................................................................................................................................................................................................................... 73
Introduction ........................................................................................................................................................................................................................... 74
Methods .................................................................................................................................................................................................................................. 75
Participants ...................................................................................................................................................................................................................... 75
Procedures ........................................................................................................................................................................................................................ 75
Results ..................................................................................................................................................................................................................................... 77
Discussion .............................................................................................................................................................................................................................. 78
Bibliography .......................................................................................................................................................................................................................... 79
VI. Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance ................................ 81
Abstract ................................................................................................................................................................................................................................... 83
Introduction ........................................................................................................................................................................................................................... 84
Methods .................................................................................................................................................................................................................................. 87
Participants ..................................................................................................................................................................................................................... 87
Apparatus ........................................................................................................................................................................................................................ 87
Procedure ......................................................................................................................................................................................................................... 89
Data and statistical analysis ....................................................................................................................................................................................... 90
Results ..................................................................................................................................................................................................................................... 92
Center of pressure displacement ................................................................................................................................................................................. 92
Frequency spectrum ...................................................................................................................................................................................................... 93
Sensory substitution ..................................................................................................................................................................................................... 94
Discussion ............................................................................................................................................................................................................................. 96
ABF efficacy in reducing sway is related to the availability of sensory information......................................................................................... 96
Attention to natural sensory information may have limited ABF efficacy in BVL subjects .............................................................................. 97
Use of ABF reduced BVL subjects’ inter-subject Variability .................................................................................................................................... 97
ABF redundancy with sensory information was higher for BVL than for control subjects .............................................................................. 98
Bibliography ........................................................................................................................................................................................................................100
X
VII. Effects of Linear vs Sigmoid Coding of Visual or Audio Biofeedback for the Control of Upright Stance ..... 105
Abstract .................................................................................................................................................................................................................................107
Introduction .........................................................................................................................................................................................................................108
Methods ................................................................................................................................................................................................................................110
Participants .................................................................................................................................................................................................................... 110
Apparatus ....................................................................................................................................................................................................................... 110
Procedure ....................................................................................................................................................................................................................... 113
Data Analysis................................................................................................................................................................................................................. 114
Results ...................................................................................................................................................................................................................................115
Effectiveness of Biofeedback ...................................................................................................................................................................................... 115
Postural Strategies ........................................................................................................................................................................................................ 115
Discussion ............................................................................................................................................................................................................................117
Bibliography .................................................................................................................................................................................................................. 120
VIII. Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance ...... 123
Abstract .................................................................................................................................................................................................................................125
Introduction .........................................................................................................................................................................................................................126
Materials & Methods .........................................................................................................................................................................................................128
Participants .................................................................................................................................................................................................................... 128
Protocol .......................................................................................................................................................................................................................... 128
ABF modalities .............................................................................................................................................................................................................. 129
Data collection and analysis ...................................................................................................................................................................................... 129
Results ...................................................................................................................................................................................................................................131
ML COP and Accelerations .......................................................................................................................................................................................... 131
Effect of ABF and learning on the correlations between ML COP and Accelerations ..................................................................................... 133
Postural Response Transfer Function ........................................................................................................................................................................ 133
Discussion ............................................................................................................................................................................................................................136
ML COP and Accelerations .......................................................................................................................................................................................... 136
Postural Responses Transfer Function ...................................................................................................................................................................... 137
Bibliography ........................................................................................................................................................................................................................140
IX. Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects .......................................... 143
Abstract .................................................................................................................................................................................................................................145
Introduction .........................................................................................................................................................................................................................146
Materials and Methods .....................................................................................................................................................................................................148
Subjects ................................................................................................................................................................................................................................148
Apparatus ....................................................................................................................................................................................................................... 148
Procedure ....................................................................................................................................................................................................................... 149
Data- and Statistical- Analysis................................................................................................................................................................................... 150
Results ...................................................................................................................................................................................................................................152
Immediate Effects of BF ............................................................................................................................................................................................... 152
Effects of Practicing Tandem Gait ............................................................................................................................................................................. 153
Effects of Practicing Tandem Gait in the Session without BF ............................................................................................................................... 153
Effects of Practicing Tandem Gait in the Session with BF ..................................................................................................................................... 154
Short-term Retention Effect of Practicing Tandem Gait........................................................................................................................................ 154
XI
Table of Contents
Discussion ............................................................................................................................................................................................................................155
Motor Learning During Tandem Gait Practice ........................................................................................................................................................ 155
Practice Sessions with and without BF ..................................................................................................................................................................... 155
Short-term Retention of Motor Learning ................................................................................................................................................................. 156
Conclusions .................................................................................................................................................................................................................... 156
Bibliography ........................................................................................................................................................................................................................158
X. Effects of Biofeedback on Postural Control and Potential Impact in Motor Rehabilitation Therapy ................ 161
Conclusions ..........................................................................................................................................................................................................................163
XII
XIII
Table of Contents
XIV
Chapter
Chapter 1
Sensory Integration and
Augmentation for the Control of
Posture
1
Sensory Integration and Augmentation for the Control of Posture
2
Chapter 1
Abstract
This first chapter is an introduction to sensorimotor integration and augmentation.
Many evidences of sensorimotor integration are presented from studies often focused on
sensorimotor illusions. These evidences support the hypothesis that the central nervous
system is continuously and unconsciously able to 1) integrate and re-weight sensorimotor
information, 2) create an internal representation of the body in space based on sensorimotor
information, and 3) re-calibrate sensorimotor information.
Some examples of movement disorders related to impaired sensorimotor integration are
also reviewed. Specifically, vestibular loss is presented as one of the pathology which could
more likely benefit from sensory information augmentation both for motor improvements
and for rehabilitation . In addition, studies aimed at demonstrating sensorimotor integration
impairments in subjects with peripheral neuropathy, Parkinson disease, and other movement
disorders are briefly reported.
Finally, sensory augmentation using biofeedback is reviewed. Many applications of
biofeedback are reported with an emphasis on biofeedback systems for postural control.
The main issues related with the design of biofeedback systems for postural control and with
the design of experimental protocols aimed at valuating biofeedback system effectiveness
are also discussed. In addition, a brief review on the application of virtual reality for postural
control assessment and improvement is presented and proposed as a highly desirable feature
for next-generation biofeedback systems for postural control.
3
Sensory Integration and Augmentation for the Control of Posture
Evidence of Sensory Integration for the Control of Posture
Control of posture is a crucial task with two main
goals: equilibrium and orientation [1]. Although we
often take the postural system for granted because it
operates primarily at a non-cognitive level, it actually
depends on a complex and active interaction among the
sensory, muscular, and nervous systems. To appreciate the
importance and the level of accuracy that this interaction
can reach in humans to maintain equilibrium and spatial
orientation, imagine a circus performer walking on a 10meters-high steel wire while juggling clubs. Now, imagine
the same juggler being suddenly in the dark, or loosing
the sensation of the wire under his/her feet and, if that
is not enough, loosing also the perception of gravity.
This example highlights is the importance of sensory
information and its integration for the achievement of
effective postural control
Photos: Michalska and Venturska street, Bratislava, June 2006.
Concurrence and Interference of Sensory Information
The control of equilibrium and orientation depends on concurrent feedback of motion
information from the vestibular, somatosensory, and visual senses. The importance of sensory
feedback is evident, for example, from the sway increase occurring when sensory information
becomes unavailable during simple quiet stance. In fact, during quiet stance, the largest
increase in postural sway occurs when somatosensory information is unavailable [2]. The next
largest increase, when vestibular information is unavailable, and the smallest, when vision
is unavailable [1;3;4]. Limitation of vestibular, visual, and somatosensory information, which
is part of the ageing process, is a major factor leading to falls in elderly people [5;6] and a
major health problem [7].
4
Chapter 1
Information from vestibular, somatosensory, and visual senses is redundant. Redundancy
of sensory information is crucial, for example, for walking in the dark (when vision is not
available), or on a compliant surface (when somatosensory information is inaccurate). In fact,
in these two situations, the central-nervous-system must rely only on two senses (vestibular
and somatosensory, and vestibular and visual, respectively). Thus, the ability to walk in the
dark or on a compliant surface does not only depend on sensory information redundancy but
also on the central-nervous-system ability to 1) evaluate and compare sensory information,
2) distinguish between reliable and unreliable information and 3) combine the sensory
information into a integrated representation of the environment. This central-nervous-system
process is known as sensory integration. The prevailing theory states that the various sources
of sensory information are integrated to form an internal model of the body that the centralnervous-system uses to plan and execute motor behaviors [1]. This internal model must be
adaptive, to accommodate changes associated with growth and development, aging, and
injury [1]. Thus, this internal model needs to be continuously recalibrated so that it can weigh
differently the motion information coming from the different senses [8;9].
One evidence of sensory interaction and remapping of an internal model of spatial
orientaton, based on vestibular sensory information, can be foreseen in the oculogyral
and audiogyral illusions. Oculogyral and audiogyral illusions are experienced by a subject
rotating with a constant angular velocity and are due to ambiguity of sensory information.
When a subject is seated on a chair rotating with a constant angular velocity both visual
and auditory spatial localization change [10;11]. In fact, if the subject is in the dark and a
head-fixed visual target is lit, the subject perceives the target as moving with his/her body,
changing the apparent position in space but leading the body as well in the direction of
the acceleration (oculogyral illusion). In a similar way, a head-fixed auditory target will be
heard by the same subject as moving in opposite direction with reference to the angular
acceleration (audiogyral illusion). Whiteside et al. (1965), [12], explained the oculogyral illusion
as due to an error in body visual localization due to the fixation of the target overriding the
vestibular nystagmus being misinterpreted as an eye deviation. Other illusions, known as
oculographic, somatogravic, and audiogravic, occur when a subject is exposed to unusual
patterns of gravitoinertial acceleration [13-15]. These illusions are due to a misinterpretation
of gravitational vertical when a subject is exposed to centrifugal acceleration in a rotation
chamber. Howard and Templeton, [16], explained these illusions as due to the inability of
the otholith to distinguish between gravitational and inertial acceleration. In 2001, another
explanation, suggesting a more complex mechanism of central remapping of sensory
localization, was also proposed by DiZio et al. [17].
As the otolith organ in the inner ear can influence the perception of the orientation of
the head and the gravito-inertial acceleration, muscle spindles have been found to influence
the perception of position of body segments [18]. In fact, by vibrating postural muscles, it is
possible to activate the muscle spindle so that illusions of motion are elicited. For example,
5
Sensory Integration and Augmentation for the Control of Posture
by vibrating the Achilles tendons of a subject, it is possible to elicit a pitch rotation at the
ankles. In addition, if a fixed, visual target is showed to the subject during the vibration, this
subject will see the target as if it were moving in the direction of apparent self-motion [19].
Another example of motion illusion comes from an experiment of Karnath et al. (1994) who
found that when neck muscles are vibrated, perception of head rotation is elicited [20]. Similar
illusions can be elicited also for other body segments [21]. In addition, if a visual or auditory
target is presented to subjects during muscle vibration, this target will be perceived to move
according to the motion illusion experienced by the subject.
The influence of haptic information on posture is extremely important and can also
induce perceptions of self-motion. Light-touch, haptic information from the index finger of a
hand touching a firm surface without any mechanical support (force applied is less than 100g)
stabilizes posture by reducing sway up to 50% in blindfolded subjects [22]. Furthermore, lighttouch improves postural stability in all subjects tested up to now such as elderly, cerebellar,
neuropathic, and labyrinthine defective subjects [23]. When the touched surface is oscillated,
subjects sway entrained to this oscillation, then trusting the haptic information more than the
other motion sensory information [24] by responding to an illusion of motion. Light-touch
information cancels out the destabilizing effect of tonic vibration reflexes in leg muscles
[25] as well as the illusion of self-displacement and airplane displacement during parabolic
flights [26].
When sensory information is ambiguous, as in the example of illusions described
above, cognitive knowledge and assumptions can influence the subject’s behavior and the
extent to which subjects perceive the illusions. For example, if subjects are aware that the
surface used for light-touch is oscillated, they may show a smaller correlated oscillation in
their sway than if they did not know about this surface motion. Cognitive knowledge is
also important, for example, to neglect the ototlith information elicited by the centrifugal
force when a sharp sudden turn occurs [27]. Internal models created by cognitive knowledge
have also been suggested to be used by the central-nervous-system to resolve ambiguity in
sensory information from the ototliths and, specifically, for distinguishing between inertial
and gravitational acceleration based also on information from the semicircular canals [28].
Tuning and Calibration of Sensory Information for Internal Model Representation
To achieve sensory integration, the central-nervous-system needs to compare
continuously the sensory information from different senses, so that matches or mismatches
among sensory information can be accurately detected and internal model of localization
conveniently tuned up and calibrated. Evidence that interaction with hands may help achieve
spatial calibration of the body comes from another illusion described by Lackner et al. (1988)
[21]. In Lackner’s experiment, a subject was holding his/her nose when the biceps brachii
muscle of the arm was vibrated. The illusion of arm extension due to the vibration, led the
6
Chapter 1
subject to feel his/her nose was elongating. This ‘Pinochio illusion’ suggests that body spatial
calibration may start from tactile interaction with the environment.
When subjects are exposed to an artificial gravity environment, head, arm, and leg
movement control, as well as locomotion, can be promptly adapted if the same motion task
is attempted repeatedly [29]. During the adaptation process, the subjects create a new model
of the environment where they integrate the new Coriolis acceleration due to the artificial
gravity. Once the adaptation is completed, Coriolis acceleration is not cognitively perceived
anymore. However, as soon as the subjects are back to natural gravity environment, the
Coriolis force, associated with the artificial gravity environment and integrated in the subjects’
internal model of the environment, is consciously perceived again as influencing the subjects’
movement until a new process of adaptation to natural gravity has been completed. Until
the adaptation process is completed, subjects show mirror-image error in movement control
compared to the ones experience in the first process of adaptation to artificial gravity. Thus,
these results suggest that the body is dynamically calibrated by its force environment, and
that movements within it feel virtually effortless.
The adaptive tuning of the body internal model of gravity must also take into account
the self-generated Coriolis forces experienced during common daily movements [30]. In fact,
whenever a natural turn-and-reach movement is performed, the simultaneous occurrence of
trunk movement and the arm forward velocity generate very high Coriolis acceleration on
the reaching arm. The preservation of reaching accuracy suggests that the central nervous
system is able to predict the Coriolis acceleration and compensate it with anticipatory forces
generated during the task.
Coriolis acceleration can be generated also by making pitch head movement during
passive rotation [31]. In this case, an illusory tumbling sensation is elicited by the Coriolis
acceleration and the subjects feel nauseous. Surprisingly, during orbital flights subjects
performing the same head rotation do not experience nausea. Further experiments on this
interesting result suggest that the lack of motion sickness during orbital flights may be due
to the lack of internally represented body displacement. However, motion sickness remains
not totally understood. Many theories have been proposed to explain motion sickness such
as sensory information conflict [32], however the only firm result is that subjects without
functioning labyrinths have not been made motion sick although several protocols have
been tried [33].
7
Sensory Integration and Augmentation for the Control of Posture
Movement Disorders and Their Relation to Sensorimotor Integration
Sensorimotor integration is the ability of using
sensory information properly for assisting motor program
execution. Whenever sensorimotor integration is impaired,
movement disorders are experienced by subjects with a
variety of different symptoms such as vertigo, dizziness,
and bradykinesia. The relationship between sensorimotor
integration and pathologies involving motor impairment
is intuitive in case of vestibular or somatosensory loss.
However, a connection between sensorimotor integration
and movement disorders has also been suggested in other
motor pathologies such as Parkinson’s disease.
Photo: A subject with Parkinson’s disease during an experiment
at the Biomechanics Laboratory of the University of Bologna in
Cesena, Italy. January 2005.
Sensory Loss and Aging
One-third to one-half of the population over age 65 reports some difficulty with balance
or ambulation [34-37]. The most common cause of impaired postural stability is the loss of
accurate and/or adequate sensory information from vestibular, somatosensory, and visual
systems [38-40]. In Europe, approximately one-third of community-dwelling adults over 65
years and fifty percent of those over 80 years fall at least once a year [41]. Twenty to thirty
percent of those who fall suffer injuries that reduce mobility and independence and increase
the risk of premature death [42-43].
Acute peripheral, vestibular loss can be caused by damage of the vestibular organ or of
the vestibular nerve and results in sensations that reflect abnormal information about head
motion [44]. The vestibular system is a bilateral organ that consists of 3 semicircular canals
and 2 otoliths on each side [45]. This complicated organ is able to provide the central nervous
system with information about the linear accelerations and angular velocities of the head in
space. The central nervous system processes this sensory information and integrates it with
the other sensory information to determine the gravitational vertical direction [1]. The most
common cause of damage of the vestibular organs (and, consequently, of bilateral vestibular
8
Chapter 1
loss) is a toxic reaction to antibiotics such as gentamicin, which selectively damages the
vestibular hair cells. Exposure to gentamicin causes bilateral vestibular loss in 3-4% of cases
[46]. Unilateral vestibular loss is common when the vestibular nerve is damaged. Specifically,
neuritis is the most common cause of unilateral vestibular loss [47]. Unilateral damage of the
vestibular nerve causes an asymmetry in the vestibular nerves firing rates. The central nervous
system interprets this asymmetry as a head rotation toward the contralesional ear. This results
in spontaneous nystagmus, with slow components in the direction of the lesioned ear and
fast in the direction of the contralesional ear.
Nystagmus is related to the vestibular organ via the vestibule-ocular reflex [48]. A
clinical measure of vestibular function is based on the vestibular-ocular reflex and observed
nystagmus is an indicator of vestibular function [49]. For example, the “head thrust test” is
based on the knowledge that when the vestibular-ocular reflex is functioning normally, the
eyes move in the direction opposite to the head movement to stabilize gaze in space [50]. The
vestibule-ocular reflex gain and phase are used to quantify vestibular loss and are normally
measured in the laboratory by recording eye movement in the dark when the subjects are
rotated in the horizontal plane. However, the vestibular-ocular reflex gain and phase during
horizontal body rotation are only indicators of the horizontal canal function and not of the
whole vestibular system. Recently, a new diagnostic method, based on the vestibular-evoked
myogenic potentials has been used to also measure saccular otolith function [51] .
Loss of vestibular function can occur slowly as in the aging process or suddenly, as in
the case of ototoxicity and neuritis described above. When vestibular loss occurs suddenly,
balance disorders are immediately evident and subjects need to go through a rehabilitation
period where they learn how to compensate for the vestibular loss before they can walk or
comfortably perform daily life motor tasks again. During this period, subjects learn how to
rely more on visual and somatosensory information to compensate the lack of vestibular loss
[52]. Classical symptoms occurring after sudden vestibular loss include: vertigo and dizziness
due to the abnormal perception of self-motion. These symptoms disappear spontaneously
over time [53]. However, some abilities as riding a bike or playing tennis may not be ever
achieved again. Even after being fully compensated, unilateral vestibular loss subjects may
show abnormal postural alignment [54], asymmetric weight distribution [55], and inability
to stand on one foot or walk with a narrow base of support [56]. In addition, whenever
unilateral or bilateral loss vestibular loss subjects are exposed to condition of altered visual
or somatosensory information, their ability to maintain balance is especially impaired [57].
Vestibular rehabilitation for subjects with unilateral deficits consists of exercises to
enhance the gain of the vestibular-ocular reflex, static and dynamic exercise with augmented
sensory information from a therapist, and activities to tolerate movement of the head and
the body [58]. Such rehabilitation therapies have been proven to be effective in helping
those with acute vestibular neuritis return to normal activities of daily living [59]. Vestibular
9
Sensory Integration and Augmentation for the Control of Posture
rehabilitation was found to be successful also in subjects with bilateral vestibular loss [60].
Vestibular rehabilitation may be useful also for elderly subjects who may not be aware of their
vestibular loss because it occurred gradually. In fact, when vestibular loss occurs gradually,
although there may be no dizziness, subjects may be unstable and fall when in an environment
requiring vestibular information for balance, i.e. in the dark on an unstable surface.
With aging, peripheral somatosensory nerve deficits also become more common. Peripheral
sensory nerve deficits lead to delay, distortion and loss of somatosensory information from
the muscles, joints, and skin which can be assessed by measuring the vibratory sensation
and ankle stretch reflex. Both vibratory sensation and the ankle stretch reflex are commonly
impaired in the elderly population [61;62]. Peripheral neuropathy can be the consequence
of several causes such as diabetes mellitus, alcoholism, nutritional deficiencies, infections,
malignancies, and autoimmune diseases [63]. Environmental and pharmaceutical agents, as
well as some hereditary factors can also lead to peripheral neuropathy. However, only in 72%
of the adults manifesting the syndromes of peripheral neuropathy, a specific cause can be
identified [64].
Some changes in the structure and function of peripheral nerves may be the result of
the aging process itself [65-68]. Peripheral neuropathy has been related to impaired balance
and falls by many studies [69-73]. Loss of sensory information from neuropathy can challenge
sensory integration in the elderly, leading to falls, or simply limiting elderly subjects’ activities
that facilitate a premature functional decline.
10
Chapter 1
Augmentation of Sensory Information for Postural Control
Sensory loss or abnormal, inadequate sensory
information from the vestibular, somatosensory and
visual senses can jeopardize the central-nervous-system
ability to control postual statbility. In this case, providing
the central-nervous-system with substitutive, artificial,
sensory information may help restore the ability to control
posture. Artificial, sensory information can also be used
to augment sensory information during rehabilitation
sessions when brain plasticity and adaptation are
crucial and depends on the extent and accuracy of the
sensory information available. Augmentation of sensory
information normally implies the use of an external device
able to provide information about body motion through
biofeedback, eventually presented in a virtual reality
environment. However, proprioception has been proven
Photo: A subjects using audio- and visual- to be augmented also by a simple sole or knee vibration
biofeedback at the Balance Disorders
Laboratory of the Oregon Health & Science ([74] and [75], respectively).
University, Portland (OR) USA, July 2005.
Overview on Biofeedback Experimentation
Biofeedback has been applied since the 50s [76] and can be defined as a process in
which a person learns to reliably influence physiological responses of two kinds: either responses
which are not ordinarily under voluntary control or responses which ordinarily are regulated but
for which regulation has broken down [77]. Since the early 60s, many studies have reported the
use of biofeedback in many areas such as instrumental conditioning of automatic nervous
system responses, psychophysiology, behavior therapy and medicine, stress research and
stress management strategies, biomedical engineering, electromyography, consciousness,
electroencephalography, cybernetics, and sports [76].
The application of biofeedback to improve postural control began in the 70s with
visual biofeedback of electromyogram, positional, or force parameters [78;79]. Studies with
11
Sensory Integration and Augmentation for the Control of Posture
electromyograms showed how subjects with sensorimotor deficits can volitionally control
single muscle activation and become more aware of muscular contraction when muscle
activation could be seen or heard [80;81]. Studies on positional and force parameters showed
how subjects could improve control of posture by actively responding to visual cues indicating
surface reactive forces provided by the biofeedback systems [82].
The neurological mechanisms underlying the effectiveness of biofeedback are still
mainly unknown. However, some hypotheses have been suggested. Two hypotheses come
from Basmajian (1982) [83] who believed that, either new pathways, or a new feedback
loops recruiting cerebral and spinal pathways already existing are used as a consequence
of exposure to biofeedback. In addition, Wolf (1983) [79], suggested that auditory and visual
stimuli from biofeedback can activate synapses that were not used before. Although not totally
understood, the effects of biofeedback seem to favor brain plasticity and, as a consequence,
shows a noticeable potential for motor rehabilitation applications.
Design of Biofeedback Systems
Three main parts essential for the design of a biofeedback system for postural control
are: 1) a sensor or an instrument able to measure some aspect of human motion, 2) a restitution
device, able to convey the biofeedback information to the subject (e.g. via the auditory, visual,
or tactile sense), and 3) some circuit or a computer able to implement a conversion algorithm
which transposes the information sensed by the sensor into a convenient activation of the
restitution device (Figure 1).
Several combinations of sensors and restitution devices, concerted by simple or complex
algorithms, have been implemented and tested to determine whether they could improve
motor control. Biofeedback systems have been proven to be effective in many areas, despite
the intrinsic difference of the wide variety of biofeedback systems designs and of their
application fields. For example, biofeedback systems were found to be effective in improving
sportive performance [84] by decreasing stress and anxiety during training in many sports
such as gymnastic [85], swimming [86], basketball [87], judo [88], archery [89], shooting [90],
and golf [91]. The use of biofeedback for improving control of posture in subjects with motor
disorders has been more oriented to provide augmented movement information of body
Sensor
Variable
Sensed
Postural Response
Coding
Subjects
Information
Coded
Biofeedback Information
Figure 1 – Diagram for the design of a biofeedback system.
12
Representation
Chapter 1
Table 1 – Published studies on the experimentation of biofeedback systems for postural control
Article
Feedback Variable
Restitution
Subjects
Application
Halpern et al., 1970
Head Position
Tactile/Mechanical
Children with Cerebral Palsy
Head Control
Hlavackha., 1973
Center of Pressure
Visual
Healthy
Standing
Harris et al.,1974
Head/Limb Position
Visual/Auditory
Athenoid Children
Head/Limb Control
Woolridge and Russel 1976
Head Position
Visual/Auditory
Children with Cerebral Palsy
Head Control
Wannstedt and Herman, 1978
Center of Pressure
Auditory
Hemiplegic
Symmetry of Standing
Walmsley et al.1981
Head Position
Auditory
Mentally Retarded Children
Head Control
Leiper et al., 1981
Head Tilt
Auditory
Children with Cerebral Palsy
Head Control
Wolf and Binder-McLeod, 1983
Muscle Activity
Auditory
Hemiplegic
Symmetry in Standing and Locomotion
Catenese and Sanford, 1984
Head Position
Visual/Auditory
Children with Spstic Quadriplegia
Childre with Diplegia or
Head Control
Bertoti and Gross, 1988
Head Position
On/Off Movie Play
Head Control
Schumway-Cook et al., 1988
Center of Pressure
Visual
Quadriplegia
Hemiparetic
Winstein et al., 1989
Weight Distribution
Visual
Hemiplegic
Symmetry in Standing and Locomotion
Domaraki et al., 1990
Head Position
Auditory
Children with disabilities
Head Control
Clarke et al., 1990
Center of Pressure
Visual
Healthy
Weight Shifting
Hamann and Krausen, 1990
Center of Pressure
Visual
Vestibular
Standing
Jobst, 1990
Center of Pressure
Visual
Brainstem and Cerebellar Lesions
Quiet Standing and Voluntary Sway
Hamman et al., 1992
Center of Gravity
Visual
Healthy
Dynamic Tasks
Barona et al., 1994
Center of Pressure
Visual
Healthy
Voluntary Sway
Edgardt, 1994
Weight Distribution
Auditory
After Stroke
Symmetry of Standing
Pertersen et al., 1996
Center of Pressure
Auditory
Stroke
Stance Perturbed by Muscle Vibration
Easton et al., 1997
Center of Pressure
Auditory
Blind
Standing
Wong et al., 1997
Weight Distribution
Visual/Auditory
Hemiplegia/Brain Injuries
Standing
Wu, 1997
Center of Gravity
Visual
Neuropathic
Perturbed Stance
Aruin et al., 2000
Knee-to-Knee Distance
Auditory
Stroke/Pelvic Instability
Gait
Rougier et al., 2002
Center of Pressure
Visual
Healthy
Standing
Tyler et al. 2003
Head Tilt
Elcetro-Tactile
Vestibular
Standing
Dault et al., 2003
Center of Pressure
Visual
Elderly and Young Healthy
Standing
Sihvonen et al., 2004
Center of Pressure
Visual
Elderly Women
Standing and Weight Shifting
Kentala et al. 2003
Trunk Tilt
Tactile
Vestibular
Standing
Dozza et al., 2005
Trunk Acceleration
Auditory
Vestibular
Standing
Heageman et al. 2005
Trunk Angular Velocity
Auditory
Vestibular
Standing and Dynamic Tasks
Chiari et al., 2005
Trunk Acceleration
Auditory
Healthy
Standing
Dozza et al., 2006
Trunk Acceleration
Visual
Healthy
Standing
Symmetry of Standing
13
Sensory Integration and Augmentation for the Control of Posture
segments than to reduce psychological stress. Table 1 summarizes some of the studies on
biofeedback aimed at improving postural control. From Table 1, it is possible to appreciate the
wide variety of biofeedback designs implemented up to now and the multitude of pathologies
they have been tested on.
As shown in Table 1, visual-biofeedback of center of pressure displacement has been the
most popular biofeedback system design for improving postural stability. Visual-biofeedback
has been extensively used for balance rehabilitation of subjects after stroke [92] in order to
reduce postural asymmetry. In addition, visual biofeedback from force plate measurements
is the only biofeedback system commercially available and diffused. In fact, systems made by
Neurocom (http://www.onbalance.com/), such as Balance Master, which are currently used for
balance training and rehabilitation, are equipped with visual-biofeedback. Recently, partly
based on work in this thesis, the interest in biofeedback design is moving from the visualbiofeedback of force plate measurements to audio- and tactile-biofeedback of inertial sensors
measurements [93;94]. This new trend in the design is driven by the intent of producing new
cost-effective and portable systems for balance training and rehabilitation. In fact, tactile
and auditory feedback do not rely on some expensive and cumbersome monitor, and do not
require power supply cabling; further, inertial sensors are one thousand times less expensive
than force plates and much smaller, portable, and sturdy.
Biofeedback is thought to have a relevant potential for rehabilitation applications
[92;95;96]. In fact, biofeedback can help subjects re-educate their motor control system during
dynamic tasks with functionally goal-oriented exercise which help the subject to explore the
environment and solve specific motor problem [95]. However, the design of such biofeedback
systems and of the most effective clinical protocol is challenging.
One of the first challenges to be faced is the determination of the variable to be fed
back. This variable should depend upon the motor control mechanism, training task, and
therapeutic goal [92]. For example, since there are studies suggesting that hand kinematics
in reaching movement is either controlled by equilibrium point shifting [104] or by creating
a virtual trajectory of end-point [98], instead of scaling muscle activity [99], a biofeedback
system for this task should use, as feedback variable, some kinematics information instead
of electromyographic information. Successful reaching also requires control of alignment
of finger-thumb opposition [100;101], as a consequence, a biofeedback system designed to
help reaching should also provide the subject with this information. The presence of more
than one relevant parameter to be controlled in the tasks presents another challenge for
the biofeedback system design. In fact, a multi-sensing, task-oriented biofeedback system
(Figure 2) should be able to feed back all information relevant for the task without distracting
or overwhelming the subject. Determining how to combine different information into one
variable, that can be fed back without too highly cognitively demand for the subject, is a
necessary feature for the design of an optimal biofeedback system. A possible help in this
14
Chapter 1
matter is the use of biomedical models to calculate and feed back several variables in realtime [102]. Other challenges for the design of a biofeedback system regard 1) the design of
an algorithm able to correctly and efficiently represent the feedback variable in a way easy
to learn and understand for the subjects [103], 2) the choice of a convenient representation
for the feedback variable that does not interfere with the task performance [103].
Experimentation of Biofeedback Systems: Protocol Design
The biofeedback systems and the protocols described in the studies reported in Table
1 are very different from each other in their designs, which were customized to encounter
the needs of different pathologies. Nevertheless, all of these studies report some beneficial
effect of the biofeedback intervention. However, in some studies [104;105] not every subjects
improved. Nevertheless, it was always possible to find a subgroup of subjects who significantly
improved their postural performance by using biofeedback. This suggests that, depending on
the different pathologies and personal characteristics, some subjects may be more suitable
than others for benefiting from biofeedback.
Although it was always possible to show some performance improvement in at least a
restricted set of the subjects exposed to biofeedback, previous studies have not quantified or
reported positive results about learning, retention, and transfer effects as a consequence of
biofeedback training. These effects are relevant because experimentation of biofeedback for
postural control has, as a major future repercussion, the use of such devices for rehabilitation.
In a rehabilitation process, it is more important for a subject to learn a task than for the same
subject to be able, in some controlled situation with some temporary artificial biofeedback
help, to reach an outstanding performance. At the same time, the goal of a rehabilitation
process is to restore the subject’s postural ability, which requires the retention of the postural
improvement achieved during the rehabilitation session. Finally, a rehabilitation exercise is
the more useful the more the improvements, acquired by a subject practicing that specific
task, transfer to other motor tasks
Despite some positive results in terms of retention and transfer effect due to biofeedback
intervention have been reported (e.g. [104;106]), many studies do not demonstrate that
biofeedback therapy leads to significant motor function recovery [92;107-109]. This lack
of conclusive results can be due to some intrinsic challenge in the experimentation of
biofeedback. One of these challenges can be foreseen in the presence of subjects that, for
personal characteristics not well understood yet, do not show any improvement (or even get
worst) when practicing with biofeedback. The presence of such subjects [104;105] affects the
experimental results, hiding the potential beneficial effect of biofeedback. Understanding the
reasons why some subjects do not benefit from biofeedback may help determining, a priori,
which subjects are suitable for benefiting from biofeedback.
Another challenge for biofeedback studies is the difficulty of determining the extent to
15
Sensory Integration and Augmentation for the Control of Posture
which improved performance with biofeedback intervention is due to biofeedback efficacy or
to the natural, spontaneous learning process induced by repetitive practice of the task [108].
A possible solution to this issue is the implementation of an experimental design in which
trials are randomized and two groups of subjects are included, so that one of the groups is
exposed to biofeedback and the other to the simple repetition of the task. The difference
between the two groups will then be a more accurate indicator of biofeedback success.
Finally, another challenge, that concerns the retention effect, is that, practicing static tasks,
such as quiet standing, seems to have less potential to transfer performance improvements to
other motor tasks than practicing dynamic tasks [108;110]. However, it has also been reported
how practicing dynamic tasks does not improve static tasks such as quiet stance [111]. This
later finding suggests that two different biofeedback therapies, one aimed to quiet stance
improvements and one aimed to dynamic tasks improvements, may be necessary in the
rehabilitation process.
In conclusion, two double-blinded, experimental designs with randomized trials [79], one
during dynamic tasks and on during static tasks, seem to be the best protocol to determine
the effectiveness and potential impact in the rehabilitation field of biofeedback systems.
However, such protocols require a larger number of subjects, longer time, and more resources
than any simple protocol aimed to describe the immediate, overall effect of biofeedback
systems on postural control.
16
1
Chapter 1
Bibliography
[1]
F. B. Horak and J.M. Macpherson, “Postural equilibrium and orientation,” in Handbook of Physiology, R.B.
Rowell and J.T. Shepherd, Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[2]
L. M. Nashner, F. O. Black, and C. Wall, III, “Adaptation to altered support and visual conditions during stance:
patients with vestibular deficits,” J Neurosci, vol. 2, no. 5, pp. 536-544, May1982.
[3]
J. M. Macpherson and J. T. Inglis, “Stance and balance following bilateral labyrinthectomy,” Prog Brain Res,
vol. 97, pp. 219-228, 1993.
[4]
R. J. Peterka and F. O. Black, “Age-related changes in human posture control: sensory organization tests,” J
Vestib Res, vol. 1, no. 1, pp. 73-85, 1990.
[5]
L. A. Brown, A. Shumway-Cook, and M. H. Woollacott, “Attentional demands and postural recovery: the
effects of aging,” J Gerontol A Biol Sci Med Sci, vol. 54, no. 4, p. M165-M171, Apr.1999.
[6]
H. C. Chen, A. B. Schultz, J. A. shton-Miller, B. Giordani, N. B. Alexander, and K. E. Guire, “Stepping over
obstacles: dividing attention impairs performance of old more than young adults,” J Gerontol A Biol Sci
Med Sci, vol. 51, no. 3, p. M116-M122, May1996.
[7]
B. E. Maki, P. J. Holliday, and G. R. Fernie, “A posture control model and balance test for the prediction of
relative postural stability,” IEEE Trans Biomed Eng, vol. 34, no. 10, pp. 797-810, Oct.1987.
[8]
F. O. Black and L. M. Nashner, “Vestibulo-spinal control differs in patients with reduced versus distorted
vestibular function,” Acta Otolaryngol, vol. 406, pp. 110-114, 1984.
[9]
A. M. Bronstein, J. D. Hood, M. A. Gresty, and C. Panagi, “Visual control of balance in cerebellar and
parkinsonian syndromes,” Brain, vol. 113 ( Pt 3), pp. 767-779, June1990.
[10] B. Clark and Graybiel A., “The effect of angular acceleration on sound localization: the audiogyral illusion,”
J Psycol, vol. 28, pp. 235-244, 1949.
[11] A. Graybiel and D.I. Hupp, “The oculogyral illusion: a form of apparent motion which may be observed
following stimulation of the semicircular canals,” Aerosp Med, vol. 3, pp. 1-12, 1946.
[12] T.C.D. Whiteside , A. Graybiel, and J.I. Niven, “Visual illusions movement,” Brain, vol. 88, pp. 193-219, 1965.
[13] B. Clark and Graybiel A., “Apparent rotation of a fixed target associated with linear acceleration in flight,”
Am J Ophthalmol, vol. 32, pp. 549-557, 1949.
[14] B. Clark and Graybiel A., “Linear acceleration and deceleration as factors influencing nonvisual orientation
during flight,” J Aviat Med, vol. 20, pp. 92-101, 1949.
[15] A. Graybiel and J.I. Niven, “The effect of a change in direction of resultant force on sound localization: the
audiogravic illusion,” J Exp Psychol, vol. 42, pp. 227-230, 1951.
[16] I.P. Howard and W.B. Templeton, Human Spatial Orientation. London: Wiley, 1966.
[17] P. DiZio, R. Held, J. R. Lackner, B. Shinn-Cunningham, and N. Durlach, “Gravitoinertial force magnitude
and direction influence head-centric auditory localization,” J Neurophysiol, vol. 85, no. 6, pp. 2455-2460,
June2001.
[18] P.C.B. Matthews, Mammalian muscle receptors and their central actions. London: Arnold, 1972.
[19] J. R. Lackner and M. S. Levine, “Changes in apparent body orientation and sensory localization induced by
vibration of postural muscles: vibratory myesthetic illusions,” Aviat Space Environ Med, vol. 50, no. 4, pp.
346-354, Apr.1979.
17
Sensory Integration and Augmentation for the Control of Posture
[20] H. O. Karnath, D. Sievering, and M. Fetter, “The interactive contribution of neck muscle proprioception and
vestibular stimulation to subjective “straight ahead” orientation in man,” Exp Brain Res, vol. 101, no. 1, pp.
140-146, 1994.
[21] J. R. Lackner, “Some proprioceptive influences on the perceptual representation of body shape and
orientation,” Brain, vol. 111, no. 2, pp. 281-297, Apr.1988.
[22] J. J. Jeka and J. R. Lackner, “Fingertip contact influences human postural control,” Exp Brain Res, vol. 100,
no. 3, pp. 495-502, 1994.
[23] J. R. Lackner, P. DiZio, J. J. Jeka, F. B. Horak, and D. Krebs, “Precisioun contact of the fingertip reduces postural
sway of individuals with bilateral vestibular loss,” Exp Brain Res, vol. 126 pp. 459-466, 1999.
[24] J. J. Jeka, G. Schoner, T. Dijkstra, P. Ribeiro, and J. R. Lackner, “Coupling of fingertip somatosensory information
to head and body sway,” Exp Brain Res, vol. 113, no. 3, pp. 475-483, Mar.1997.
[25] J. R. Lackner and P. A. DiZio, “Aspects of body self-calibration,” Trends Cogn Sci, vol. 4, no. 7, pp. 279-288,
July2000.
[26] J. R. Lackner and P. DiZio, “Multisensory, cognitive, and motor influences on human spatial orientation in
weightlessness,” J Vestib Res, vol. 3, no. 3, pp. 361-372, 1993.
[27] T. Imai, S. T. Moore, T. Raphan, and B. Cohen, “Interaction of the body, head, and eyes during walking and
turning,” Exp Brain Res, vol. 136, no. 1, pp. 1-18, Jan.2001.
[28] D. M. Merfeld, L. Zupan, and R. J. Peterka, “Humans use internal models to estimate gravity and linear
acceleration,” Nature, vol. 398, no. 6728, pp. 615-618, Apr.1999.
[29] P. DiZio and J. R. Lackner, “Motor adaptation to Coriolis force perturbations of reaching movements:
endpoint but not trajectory adaptation transfers to the nonexposed arm,” J Neurophysiol, vol. 74, no. 4,
pp. 1787-1792, Oct.1995.
[30] P. Pigeon, S. B. Bortolami, P. DiZio, and J. R. Lackner, “Coordinated turn-and-reach movements. I. Anticipatory
compensation for self-generated coriolis and interaction torques,” J Neurophysiol, vol. 89, no. 1, pp. 276289, Jan.2003.
[31] W. H. Johnson, R. A. Stubbs, G. F. Kelk, and W. R. Franks, “Stimulus required to produce motion sickness. I.
Preliminary report dealing with importance of head movements,” J Aviat Med, vol. 22, no. 5, pp. 365-374,
Oct.1951.
[32] J. T. Reason, “Motion sickness: a special case of sensory rearrangement,” Adv Sci, vol. 26, no. 130, pp. 386393, June1970.
[33] A. Graybiel, “Structural elements in the concept of motion sickness,” Aerosp Med, vol. 40, no. 4, pp. 351-367,
Apr.1969.
[34] C. Coogler, “Falls and imbalance,” Rehab Management, p.53, April/May 1992.
[35] J. M. Hausdorff, D. A. Rios, and H. K. Edelberg, “Gait variability and fall risk in community-living older adults:
a 1-year prospective study,” Arch Phys Med Rehabil, vol. 82, no. 8, pp. 1050-1056, Aug.2001.
[36]
“Healthy People 2010 Understanding and Improving Health,” Washington D.C., 2000.
[37] Robbins A.S, Rubinstein L.Z., Josephson K.R., Schulman B.L., Osterweil D., and Fine G., “Predictors of falls
among elderly people. Results of two population-based studies,” Archives of Internal Medicine, vol. 149,
no. 7, pp. 1628-1633, 1989.
[38] F. B. Horak, C. L. Shupert, and A. Mirka, “Components of postural dyscontrol in the elderly: a review,”
Neurobiol Aging, vol. 10, no. 6, pp. 727-738, Nov.1989.
[39] S. R. Lord and J. A. Ward, “Age-associated differences in sensori-motor function and balance in community
dwelling women,” Age Ageing, vol. 23, no. 6, pp. 452-460, Nov.1994.
[40] H. Corriveau, F. Prince, R. Hebert, M. Raiche, D. Tessier, P. Maheux, and J. L. Ardilouze, “Evaluation of postural
stability in elderly with diabetic neuropathy,” Diabetes Care, vol. 23, no. 8, pp. 1187-1191, Aug.2000.
[42] M. Halter, “Falls in the older population: a pilot study to assess those individuals who are attended to by
the London Ambulance Service as a result of a fall but are not conveyed to an accident and emergency
department”., London Ambulance Service NHS Trust, 2000.
[43] M.E. Tinetti, M. Speechley, “Prevention of falls among the elderly,” New England journal of medicine, vol.
18
Chapter 1
320 pp. 1055-1059, 1989.
[43] C. Freeman, “Quality improvement for people with hip fracture: experience from a multi-site audit,” Quality
and safety in health care, vol. 11, pp. 239-245, 2002.
[44] L. B. Minor, “Physiological principles of vestibular function on earth and in space,” Otolaryngol Head Neck
Surg, vol. 118, no. 3, pp. S5-15, Mar.1998.
[45] V. J. Wilson and G. Melvill Jones, Mammalian Vestibular Physiology. New York: Plenum, 1979.
[46] G. Kahlmeter and J. I. Dahlager, “Aminoglycoside toxicity - a review of clinical studies published between
1975 and 1982,” J Antimicrob Chemother, suppl. A, vol. 13, pp. 9-22, Jan.1984.
[47] C. W. Cooper, “Vestibular neuronitis: a review of a common cause of vertigo in general practice,” Br J Gen
Pract, vol. 43, no. 369, pp. 164-167, Apr.1993.
[48] B. Cohen, S. Wearne, M. Dai, and T. Raphan, “Spatial orientation of the angular vestibulo-ocular reflex,” J
Vestib Res, vol. 9, no. 3, pp. 163-172, 1999.
[49] M. C. Schubert and L. B. Minor, “Vestibulo-ocular physiology underlying vestibular hypofunction,” Phys Ther,
vol. 84, no. 4, pp. 373-385, Apr.2004.
[50] E. F. I. Miller and Graybiel A., “Counterrolling of the human eyes produced by head tilt with respect to
gravity,” Acta Otolaryngol., vol. 54, pp. 479-501, 1962.
[51] J. G. Colebatch and G. M. Halmagyi, “Vestibular evoked potentials in human neck muscles before and after
unilateral vestibular deafferentation,” Neurology, vol. 42, no. 8, pp. 1635-1636, Aug.1992.
[52] F. B. Horak and C. L. Shupert, “Role of the vestibular system in postural control,” in Vestibular Rehabilitation.
S. J. Herdman, Ed. Philadelphia: F. A. Davis Company, pp. 22-46, 1994.
[53] H. Flohr, H. Beeinhold, and W. Abeln, “Concepts of Vestibular Compensation,” in Lesion Induced Neuronal
Plasticity in Sensorimotor Systems. Precht, Ed. Berlin, Germany: Springer-Verlag, pp. 153-172, 1981.
[54] A. Shumway-Cook and F. B. Horak, “Rehabilitation strategies for patients with vestibular deficits,” Neurol
Clin, vol. 8, no. 2, pp. 441-457, May1990.
[55] S. Takemori, M. Ida, and H. Umezu, “Vestibular training after sudden loss of vestibular functions,” J
Otorhinolaryngol Relat Spec., vol. 47, no. 2, pp. 76-83, 1985.
[56] H. C. Diener, F. Horak, G. Stelmach, B. Guschlbauer, and J. Dichgans, “Direction and amplitude precuing has
no effect on automatic posture responses,” Exp Brain Res, vol. 84, no. 1, pp. 219-223, 1991.
[57] A. Shumway-Cook and F. B. Horak, “Assessing the influence of sensory interaction of balance. Suggestion
from the field,” Phys Ther, vol. 66, no. 10, pp. 1548-1550, Oct.1986.
[58] S. L. Whitney and M. M. Rossi, “Efficacy of vestibular rehabilitation,” Otolaryngol Clin North Am, vol. 33, no.
3, pp. 659-672, June2000.
[59] K. Tokumasu, A. Fujino, and H. Noguchi, “Prolonged dysequilibrium in three cases with vestibular neuronitis:
efficacy of vestibular rehabilitation,” Acta Otolaryngol, vol. 503, pp. 39-46, 1993.
[60] D. E. Krebs, K. M. Gill-Body, P. O. Riley, and S. W. Parker, “Double-blind, placebo-controlled trial of rehabilitation
for bilateral vestibular hypofunction: preliminary report,” Otolaryngol Head Neck Surg, vol. 109, no. 4, pp.
735-741, Oct.1993.
[61] H.H. Shaumberg, P.S. Spencer, and J. Ochoa, “The aging human peripheral nervous system,” in The neurology
of aging. R. Katzaman and R.D. Terry, Eds. Philadelpia: Davis, 1983.
[62] B. Jordan and J.L. Cummings, “Mental status and neurologic examination in the elderly,” in Principles of
geriatric medicine and gerontology. W.R. Hazzard, J.P. Blass, and W.H. Ettinger, Eds. New York: McGraw-Hill,
1999.
[63] J. W. Mold, S. K. Vesely, B. A. Keyl, J. B. Schenk, and M. Roberts, “The prevalence, predictors, and consequences
of peripheral sensory neuropathy in older patients,” J. Am. Board Fam. Pract., vol. 17, no. 5, pp. 309-318,
Sept.2004.
[64] J. George and J. A. Twomey, “Causes of polyneuropathy in the elderly,” Age Ageing, vol. 15, no. 4, pp. 247249, July1986.
[65] L. J. Dorfman and T. M. Bosley, “Age-related changes in peripheral and central nerve conduction in man,”
19
Sensory Integration and Augmentation for the Control of Posture
Neurology, vol. 29, no. 1, pp. 38-44, Jan.1979.
[66] P. Bouche, F. Cattelin, O. Saint-Jean, J. M. Leger, S. Queslati, D. Guez, A. Moulonguet, Y. Brault, J. P. Aquino,
and P. Simunek, “Clinical and electrophysiological study of the peripheral nervous system in the elderly,”
J Neurol, vol. 240, no. 5, pp. 263-268, May1993.
[67] P. K. Taylor, “Non-linear effects of age on nerve conduction in adults,” J Neurol Sci, vol. 66, no. 2-3, pp. 223234, Nov.1984.
[68] J. M. Jacobs and S. Love, “Qualitative and quantitative morphology of human sural nerve at different ages,”
Brain, vol. 108, no. Pt 4, pp. 897-924, Dec.1985.
[69] J. K. Richardson, “The clinical identification of peripheral neuropathy among older persons,” Arch Phys Med
Rehabil, vol. 83, no. 11, pp. 1553-1558, Nov.2002.
[70] K. Koski, H. Luukinen, P. Laippala, and S. L. Kivela, “Risk factors for major injurious falls among the homedwelling elderly by functional abilities. A prospective population-based study,” Gerontology, vol. 44, no.
4, pp. 232-238, 1998.
[71] H. Luukinen, K. Koski, P. Laippala, and S. L. Kivela, “Predictors for recurrent falls among the home-dwelling
elderly,” Scand J Prim Health Care, vol. 13, no. 4, pp. 294-299, Dec.1995.
[72] G. S. Sorock and D. M. Labiner, “Peripheral neuromuscular dysfunction and falls in an elderly cohort,” Am
J Epidemiol, vol. 136, no. 5, pp. 584-591, Sept.1992.
[73] J. K. Richardson and E. A. Hurvitz, “Peripheral neuropathy: a true risk factor for falls,” J Gerontol A Biol Sci
Med Sci, vol. 50, no. 4, p. M211-M215, July1995.
[74] A. A. Priplata, B. L. Patritti, J. B. Niemi, R. Hughes, D. C. Gravelle, L. A. Lipsitz, A. Veves, J. Stein, P. Bonato, and J.
J. Collins, “Noise-enhanced balance control in patients with diabetes and patients with stroke,” Ann Neurol,
vol. 59, no. 1, pp. 4-12, Jan.2006.
[75] D. C. Gravelle, C. A. Laughton, N. T. Dhruv, K. D. Katdare, J. B. Niemi, L. A. Lipsitz, and J. J. Collins, “Noiseenhanced balance control in older adults,” Neuroreport, vol. 13, no. 15, pp. 1853-1856, Oct.2002.
[76] M.S. Schwartz and F. Andrasik, “Biofeedback in Neuromuscular Reeducation and Gait Training,” in
Biofeedback - a Practitioners’s Guide, 3rd ed New York: Guilford Press, pp. 485-513, 2003.
[77] E. B. Blanchard and Epstein L.H., A biofeedback primer. Reading, MA: Addison-Wesley, 1978.
[78] C. K. Fernando and J. V. Basmajian, “Biofeedback in physical medicine and rehabilitation,” Biofeedback Self
Regul, vol. 3, no. 4, pp. 435-455, Dec.1978.
[79] S. L. Wolf, “Electromyographic biofeedback applications to stroke patients. A critical review,” Phys. Ther.,
vol. 63, no. 9, pp. 1448-1459, Sept.1983.
[80] J. V. Basmajian, “Research foundations of EMG biofeedback in rehabilitation,” Biofeedback Self Regul, vol.
13, no. 4, pp. 275-298, Dec.1988.
[81] A. J. Prevo, S. L. Visser, and T. W. Vogelaar, “Effect of EMG feedback on paretic muscles and abnormal cocontraction in the hemiplegic arm, compared with conventional physical therapy,” Scand J Rehabil Med,
vol. 14, no. 3, pp. 121-131, 1982.
[82] S. Moore and M. Woollacott, “The use of biofeedback to improve postural stability,” Phys Ther Practice, vol.
2, pp. 1-19, 1993.
[83] J. V. Basmajian, “Clinical use of biofeedback in rehabilitation,” Psychosomatics, vol. 23, pp. 67-73, 1982.
[84] B. Blumenstein, “Biofeedback applications in sport and exercise: research findings,” in Brain and body in
sport and exercise. B. Blumenstein, M. Bar-Eli, and G. Tenenbaum, Eds. Chichester, UK: John Wiley & Sons,
2002, pp. 37-54.
[85] E. Peper and A. Schmid, “The use of electrodermal biofeedback for peak performance training,” Somatics,
vol. 4, pp. 16-18, 1983.
[86] B. Blumenstein, M. Bar-Eli, and G. Tenenbaum, “The augmenting role of biofeedback: effects of autogenic,
imagery and music training on physiological indices and athletic performance,” J Sports Sci, vol. 13, no. 4,
pp. 343-354, Aug.1995.
[87] D. J. De Witt, “Cognitive and biofeedback trainig for stress reduction with university athletes,” J. Sport
Psychology, vol. 3, pp. 271-282, 1980.
20
Chapter 1
[88] B. Blumenstein, M. Bar-Eli, and G. Tenenbaum, “A five-speps approach to mental training incorporating
biofeedback,” The Sport Psychologist, vol. 11, pp. 440-453, 1997.
[89] D. M. Landers, S. J. Petruzzello, W. Salazar, D. J. Crews, K. A. Kubitz, T. L. Gannon, and M. Han, “The influence
of electrocortical biofeedback on performance in pre-elite archers,” Med Sci.Sports Exerc, vol. 23, no. 1, pp.
123-129, Jan.1991.
[90] B. D. Hatfield, D. M. Landers, and W. J. Ray, “Cognitive processes during self-paced motor performance: an
electroencephalografic study of elite rifle shooters,” J Aeronautical Society, vol. 6, pp. 42-58, 1984.
[91] D. J. Crews and D. M. Landers, “Electroencephalographic measures of attentional patterns prior to the golf
putt,” Med. Sci. Sports Exerc., vol. 25, no. 1, pp. 116-126, Jan.1993.
[92] D. S. Nichols, “Balance retraining after stroke using force platform biofeedback,” Phys Ther, vol. 77, no. 5,
pp. 553-558, May1997.
[93] C. Wall, III, D. M. Merfeld, S. D. Rauch, and F. O. Black, “Vestibular prostheses: the engineering and biomedical
issues,” J Vestib Res, vol. 12, no. 2-3, pp. 95-113, 2002.
[94] L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[95] J. P. Wann and J. D. Turnbull, “Motor skill learning in cerebral palsy: movement, action and computerenhanced therapy,” Baillieres Clin Neurol, vol. 2, no. 1, pp. 15-28, Apr.1993.
[96] L. Bradley, B. B. Hart, S. Mandana, K. Flowers, M. Riches, and P. Sanderson, “Electromyographic biofeedback
for gait training after stroke,” Clin Rehabil, vol. 12, no. 1, pp. 11-22, Feb.1998.
[97] W. Abend, E. Bizzi, and P. Morasso, “Human arm trajectory formation,” Brain, vol. 105, no. Pt 2, pp. 331-348,
June1982.
[98] E. Bizzi, N. Accornero, W. Chapple, and N. Hogan, “Posture control and trajectory formation during arm
movement,” J Neurosci, vol. 4, no. 11, pp. 2738-2744, Nov.1984.
[99] S. W. Keele, “Movement control in skilled movement,” Psychological bulletin, vol. 70, pp. 387-403, 1968.
[100] Y. Paulignan, V. G. Frak, I. Toni, and M. Jeannerod, “Influence of object position and size on human prehension
movements,” Exp Brain Res, vol. 114, no. 2, pp. 226-234, Apr.1997.
[101] M. Gentilucci, E. Daprati, M. Gangitano, M. C. Saetti, and I. Toni, “On orienting the hand to reach and grasp
an object,” Neuroreport., vol. 7, no. 2, pp. 589-592, Jan.1996.
[102] D. Jack, R. Boian, A. S. Merians, M. Tremaine, G. C. Burdea, S. V. Adamovich, M. Recce, and H. Poizner, “Virtual
reality-enhanced stroke rehabilitation,” IEEE Trans Neural Syst Rehabil Eng, vol. 9, no. 3, pp. 308-318,
Sept.2001.
[103] E. Jovanov, K. Wegner, V. Radivojevic, D. Starcevic, M. S. Quinn, and D. B. Karron, “Tactical audio and acoustic
rendering in biomedical applications,” IEEE Trans Inf Technol Biomed, vol. 3, no. 2, pp. 109-118, June1999.
[104] C. P. Wooldridge and G. Russell, “Head position training with the cerebral palsied child: an application of
biofeedback techniques,” Arch Phys Med Rehabil, vol. 57, no. 9, pp. 407-414, Sept.1976.
[105] A. A. Catanese and D. A. Sandford, “Head-position training through biofeedback: prosthetic or cure?,” Dev
Med Child Neurol, vol. 26, no. 3, pp. 369-374, June1984.
[106] R. E. Schleenbaker and A. G. Mainous, III, “Electromyographic biofeedback for neuromuscular reeducation
in the hemiplegic stroke patient: a meta-analysis,” Arch Phys Med Rehabil, vol. 74, no. 12, pp. 1301-1304,
Dec.1993.
[107] M. Glanz, S. Klawansky, W. Stason, C. Berkey, N. Shah, H. Phan, and T. C. Chalmers, “Biofeedback therapy in
poststroke rehabilitation: a meta-analysis of the randomized controlled trials,” Arch Phys Med Rehabil, vol.
76, no. 6, pp. 508-515, June1995.
[108] J. D. Moreland, M. A. Thomson, and A. R. Fuoco, “Electromyographic biofeedback to improve lower extremity
function after stroke: a meta-analysis,” Arch Phys Med Rehabil, vol. 79, no. 2, pp. 134-140, Feb.1998.
[109] S. L. Wolf, P. A. Catlin, S. Blanton, J. Edelman, N. Lehrer, and D. Schroeder, “Overcoming limitations in elbow
movement in the presence of antagonist hyperactivity,” Phys Ther, vol. 74, no. 9, pp. 826-835, Sept.1994.
[110] G. R. Colborne, S. J. Olney, and M. P. Griffin, “Feedback of ankle joint angle and soleus electromyography in
21
Sensory Integration and Augmentation for the Control of Posture
the rehabilitation of hemiplegic gait,” Arch Phys Med Rehabil, vol. 74, no. 10, pp. 1100-1106, Oct.1993.
[111] R. Barona, E. Zapater, J. Montalt, M. Armengot, and J. Basterra, “Efectos del entrenamiento en plataformas
dinamométricas por medio de feedback visual en el mantenimiento del equilibrio en sujetos normales,”
Acta Otorrinolaring Esp, vol. 45, pp. 161-163, 1994. [159]
[112] H. Huang, S. L. Wolf, and J. He, “Recent developments in biofeedback for neuromotor rehabilitation,” J
Neuroengineering Rehabil, vol. 3:11., p. 11, June2006.
22
Chapter 1
23
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
24
Chapter
Chapter 2
Audio-Biofeedback for Balance
Improvements: an AccelerometryBased System
Most of the content of this chapter has been published in: L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D.
Giansanti, “Audio-biofeedback for balance improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52,
no. 12, pp. 2108-2111, Dec.2005.
25
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
26
Chapter 2
Abstract
This paper introduces a prototype audio-biofeedback system for balance improvement
through the sonification of trunk kinematic information. In tests of this system, normal healthy
subjects performed several trials in which they stood quietly in three sensory conditions
while wearing an accelerometric sensory unit and headphones. The audio-biofeedback system
converted in real-time the two-dimensional horizontal trunk accelerations into a stereo sound
by modulating its frequency, level, and left/right balance.
Preliminary results showed that subjects improved balance using this audio-biofeedback
system and that this improvement was greater the more that balance was challenged by
absent or unreliable sensory cues. In addition, high correlations were found between the
center of pressure displacement and trunk acceleration, suggesting accelerometers may be
useful for quantifying standing balance.
27
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
Introduction
A complex interplay between feedback and feedforward control results in the ability
of the human body to stabilize and to maintain balance in an upright stance and during
movement [1]. Visual, vestibular, and somatosensory receptors provide the central nervous
system (CNS) with sensory information about body orientation and motion in space in order
to maintain balance. Balance deficits are frequently associated with diseases, disorders, and
conditions in which there is either incomplete environmental information supplied to the
CNS by the senses, such as in vestibular disorders, or a deterioration of the circuitry of the
CNS, such as in stroke or Parkinson’s disease. One approach to improving balance, which has
been widely used in physical therapy and rehabilitation, involves feeding back to the CNS
supplementary environmental information about body motion. This supplemental information
may be coming from artificial sensors, a therapist, or laboratory equipments [2], [3].
In the past few years, increases in the speed of microprocessors, advances in miniature
devices, and a growing interest in noninvasive patient monitoring and management have
stimulated the development of real-time portable biomedical systems that are compact and
have low cost [4], [5]. One promising application of such systems is biofeedback, which can
be used to enhance human perception of automatic biological processes, such as movement
and balance [6], [7]. Recently, Giansanti et al. [8] developed a portable sensor consisting of
three accelerometers and three gyroscopes that estimate three-dimensional (3-D) kinematic
information of a body segment. This sensor became part of the audio-biofeedback (ABF)
prototype device presented in this paper. In this paper, we will 1) describe the architecture and
the functioning principle of this ABF system, and 2) present the results of a preliminary study
that tested the hypothesis that ABF benefits normal, healthy subjects most when sensory
information is partly compromised.
28
Chapter 2
Materials and Methods
In this study, we used our customized ABF device, and a force plate (AMTI OR6-6,
Watertown, MA) to estimate body sway by means of center of pressure (COP) data. The ABF
device has three major component: 1) a sensory unit, 2) a processing unit, and 3) an audiooutput unit. The force plate was used for cross-validation and is not a component of the ABF
system.
Sensory Unit
The prototype uses a portable sensory unit described elsewhere [8], weighing about 100
grams. Briefly, the sensory unit incorporates a cell with two linear uni-axial accelerometers
(3031-Euro Sensor, UK), packaged into a 7.5x7.5x3.5 mm3 module. The accelerometers have
the following specifications: range = ±2g, sensitivity = 3 mV/g, linearity = 0.08 %FS, frequency
response = 0-350 Hz, and peak-to-peak noise = 0.15 mg over the entire bandwidth. The
accelerometers are aligned with an orthogonal reference frame rigid with the cell, and they
measure the linear accelerations of the trunk in the anterior-posterior (AP) and medial-lateral
(ML) directions.
The performance of the accelerometers in the sensory unit was previously evaluated
during several postural tasks related to activities of daily living. Results from these studies
were compared with simultaneous recordings from an optoelectronic stereo-photogrammetric
system. The sensory unit performed at a maximum error of about 10 -4 g in horizontal
accelerations.
Processing Unit
The acceleration outputs of the sensory unit are analog-to-digital converted by a DAQ
board (NI-6024E, National Instruments, Austin, TX), and processed on a Toshiba laptop computer
(CPU: Intel Celeron 2.0 GHz) running Matlab Data Acquisition Toolbox (Mathworks, Natick, MA).
Digital processing in the laptop computes the proper frequency, level, and left/right (L/R)
balance of the audio output signal. The laptop also digitizes and stores additional signals
for future analysis. Signals such as the complete 3-D linear and angular trunk kinematics are
recorded by the portable sensor. Ground reaction forces and moments are recorded by the
force plate on which subject stands.
29
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
Audio Output Unit
After digital processing in the laptop, the DAQ board converts the audio output signal
into a binaural, synthetic feedback signal flowing through a common audio amplifier (Fostex
PH-5, Japan) into headphones (Philips SBC HP-140, The Netherlands) that the user wears.
Algorithm for ABF Sound Generation
The algorithm for ABF sound generation is designed to convey spatial information about
the horizontal movements of the user’s trunk to the headphones by means of sinusoidal tones.
The audio signal maps AP and ML accelerations into stereo sound modulated in frequency,
level, and L/R balance. The ABF system uses independently modulated right and left output
channels for sound representation, with a 20-Hz refresh rate.
To avoid an overload of sensory information presented to the user, the ABF evaluates
a region around a user’s natural stance posture where subtle, spontaneous sway with small
accelerations always occurs, even in normal, healthy individuals who are able to use all of
their senses available. We refer to this region as the reference region (RR). The RR is considered
to be the area in which an individual sways while standing but still does not need any extra
information to stabilize upright posture. When swaying outside this region, an individual
receives sensory feedback to correct sway to within the RR in order to stabilize upright
posture. Our goal is to help an individual correct sway to within the RR, therefore, stabilizing
upright posture by using ABF.
Level [mV]
A
Frequency [Hz]
B
55
5
RR
SR
C
2
Level [mV]
55
AP acceleration [mm/s ]
RR
SR
AP acceleration [mm/s ]
2
L/R Balance
D
1
0.5
5
RR ML acceleration [mm/s2]
SR
0
RR ML acceleration [mm/s2]
SR
Figure 1 – (a) Level and (b) frequency modulation functions
based on anterior/posterior (AP) accelerations. (c) Level and
(d) left and right balance modulation functions based on ML
accelerations.
30
The size of an individual’s RR
is subject-specific and is defined as
a function of the person’s height.
To calculate RR, we use an inverted
pendulum model and assume, as RR
threshold, an acceleration that keeps
the angular sway within ±1 deg from
initial position [9]. Because forward
sway is usually larger than in any other
direction, we use the value obtained
from the inverted pendulum model
to set the anterior threshold of the RR,
and we empirically assign a coefficient
of 2/3 to obtain the posterior, left, and
right thresholds.
To determine an upper bound
for acceptable accelerations to help
Chapter 2
an individual stabilize upright posture, the ABF processes the limits of the stability of each
user, which we term the safety region (SR). The SR is defined as a function of the user’s feet
dimensions. Since a body can maintain its balance while standing in static conditions if its
center of mass (COM) projection falls inside its support base, we can roughly estimate an
individual’s limits of stability, i.e. his/her SR as a function of the size of the support base. The
borders of the SR represent the maximum acceleration of the user’s trunk just before the COM
projects outside the base of support delimited by the dimensions of the person’s feet.
The ABF device is designed to take advantage of human hearing, which recognizes
differences in sound frequency more easily if a reference sound is given for comparison [10].
In our ABF system, when a user sways within the RR, the ABF sends a stereo, low-volume (a
few dBs above the hearing threshold), pure tone (f0 = 400 Hz) almost equivalent to the G
above the middle C to the user via the headphones. However, when the user sways outside
the RR, the ABF sends different tones which signal to the user that sway needs correcting
and how to correct it.
We used the interval between SR and RR to design the dynamic range of the audio
output, as shown in . 1. The sigmoid functions of Fig. 1(a) and Fig. 1(c) represent the coding
laws for the generation of the sound level (expressed as input voltage to the headphones)
based on the accelerations recorded along the AP and ML directions, respectively.
The general equation of the sigmoid functions is
(1)
where a = max(aAP; aML) is the maximum dimensionless ratio between the actual amount
of acceleration exceeding the RR threshold and the RR-to-SR acceleration excursion in AP and
ML; k = 3 for the anterior direction; k = 2.5 for the posterior and ML directions; b = 0.3; L0 =
50mV rms defines the sound range; and c = 5mV rms sets the minimum signal level in the
headphones. The consequent range of the output level may be as wide as 20 dB-SPL. The
frequency modulation associated with AP acceleration follows the piecewise linear law [see
(2)] shown in Fig. 1(b).
(2)
where m = 250 Hz outside RR backward, m = 0Hz inside RR, and m = 600 Hz outside RR
forward. The amplitude and sign of the ML acceleration regulate the L/R balance between
the audio channels [see Fig. 1(d)]. Given the weighting function
31
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
(3)
where sgn(.) represents the signum function, the left and right levels are computed as
(4)
When the subject’s ML sway is inside the RR, aML = 0 and w = 0. Hence, the L/R levels
are equal. Fig. 2 shows an example of the ABF variables during a representative experiment
as processed in real-time by the computer, based on (1)–(4).
B
AP direction
100
ML acceleration [mm/s2]
-50
50
0
-50
-100
1
2
3
4
time [s]
5
6
7
1
2
3
4
time [s]
5
6
7
1
2
3
4
time [s]
5
6
7
50
AP level [mV]
ML level [mV]
45
5
5
1
2
3
4
time [s]
5
6
7
L/R balance (channel R)
800
AP frequency [Hz]
RR
RR
0
ML direction
100
SR
50
SR
AP acceleration [mm/s2]
A
100
600
400
200
1
2
3
4
time [s]
5
6
7
1
0.5
0
1
2
3
4
time [s]
5
6
7
Figure 2 - Example of sound coding for part of a representative trial. (A) AP raw acceleration (top) is
converted into sound level (middle) by (1) and into sound frequency (bottom) by (2). (B) ML raw acceleration
(top) is converted into sound level (middle) by (1) and into left and right balance (bottom) by (3) and (4).
32
Chapter 2
Experimental Protocol and Results
Several pilot experiments were performed to develop the ABF system. Critical steps in
the design phase involved 1) defining the RR, the SR, and the functions to relate sound and
body movements and 2) developing the digital sound generation process.
The validity and usefulness of the ABF system were evaluated in a preliminary experiment
in which nine normal, healthy subjects used the ABF device to maintain balance while standing
quietly. Their mean age, height, and weight were 55 (33-71 years), 167 cm (151-180 cm), and
73 kg (65-86 kg), respectively.
Each subject performed 13 trials (60 s each) with ABF while standing quietly on a force
plate, in three different conditions: five trials with eyes closed (EC), five trials with eyes open
and with foam under feet (EOF), and three trials with eyes closed and with foam under feet
(ECF). Each subject also performed the same trials without ABF, for a total of 26 trials. The eyesclosed conditions eliminated visual information. The foam-under-feet conditions, achieved
by covering the force plate with a 10-cm-thick, medium density Temper foam (Kees Goebel
Medical, Inc, Hamilton, OH), made somatosensory information from the surface unreliable.
The order of the trials was randomized.
For all trials, the sensory unit was mounted on the subject’s back, as close as possible
to the body COM by taking the subject’s navel at the height of L5 as a reference. The first
10 s of each trial, regardless of sensory condition, were used for hardware re-calibration to
reduce the effect of any possible drift of the sensors. A two-dimensional bubble placed on
the sensory unit helped correct the alignment of the sensor. To maximize the repeatability of
the procedure, the same experimenter mounted the sensory unit on all subjects.
For the trials with ABF, the subjects were instructed to keep the reference sound as
constant as possible, thus indicating that postural sway was maintained within the RR. Before
recording the trials, each subject performed one practice trial 1-min long to experience the
relation between sound and movement, and to gain confidence with the ABF system.
During each trial, COP data from the force plate and accelerations from the portable
sensor were recorded at a 100-Hz sample rate. Comparisons among the three sensory
conditions concentrated on the following five COP variables: root mean square distance (RMS),
mean velocity (MV), frequency containing 95% of the power (F95%), frequency dispersion
(FD), and direction of maximum sway variability (|90-Mdir|) [11]. The same five variables were
33
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
computed also from the acceleration signals.
Our initial analysis assessed the relationship between the COP displacement and trunk
acceleration. Thus, we performed a correlation analysis in the time domain and a coherence
analysis in the frequency domain between the two signals in all three of the sensory
conditions, with and without ABF. We also performed a correlation analysis between the five
COP and five acceleration variables. Not surprisingly, COP displacement and trunk acceleration
were largely mutually dependent (Fig. 3) [12]. As expected for an inverted pendulum model
of postural sway, the correlation coefficients found between the COP and trunk acceleration
signals along the AP and ML axes were high in all three sensory conditions (0.7 < r < 0.9).
Regarding the effect of ABF on correlations, the change in correlation coefficient r was largely
negligible, except in the ECF condition, where ABF reduced r slightly (r = 0.87 ±0:02 without
ABF, r = 0.78 ±0.03 with ABF) but systematically in both the AP and ML directions.
The coherence between COP displacement and trunk acceleration along the AP and ML
axes was high (>0.8) for frequencies below 1 Hz, peaking at 0.5 Hz. This finding is in agreement
with the low-pass nature of the biomechanical filter that relates trunk (and body) motion and
the location of the COP [1].
COP displacement RMS and acceleration RMS were the variables with the strongest
correlations (r = 0.74), while the other parameters had lower correlations: MV: r = 0.36, F95%:
r = 0.36, FD: r = 0.62, and |90-Mdir|: r = 0.50.
ML movement
AP movement
Fig. 4 shows the percentage change due to ABF observed in the COP-based parameters
in all three sensory conditions. Using ABF in the EC conditions, all nine subjects swayed less, as
reflected by the reduction of COP displacement RMS (statistically significant in EC, p < 0.05; in
ECF, p < 0.01). In addition, using ABF, most of the subjects applied more postural corrections
to their sway, as shown by the increase in MV (statistically significant in EOF, p < 0.01)
and F95% (consistently statistically
10 [mm]
[mm/s2] 100
AP acceleration
significant across conditions, p < 0.01).
50
5
0
ABF had no clear influence on FD and
0
-50
-5
|90-Mdir|. The more challenging the
AP COP displacement
-100
-10
sensory condition, the more that ABF
1
3
0
2
4
5
6
7
8
9
10
affected both stability and postural
[mm/s2] 100
10 [mm]
ML acceleration
corrections. In fact, ABF benefited
50
5
subjects’ maintenance of stance within
0
0
-50
the RR the most in the ECF condition.
-5
ML COP displacement
-100
-10
The corresponding values of the COP
3
4
5
6
7
8
9
10
0
1
2
Time [s]
parameters, expressed as mean (±SD),
were: without ABF: RMS = 14.8 (±3.9)
Figure 3 - Correlations between COP and trunk acceleration
signals. AP (top) and ML (bottom) components of COP
mm, MV = 27.7 (±11.3) mm/s, F95% =
displacement (black line) and acceleration (gray line) for a
part of a representative trial are shown.
1.59 (±0.18) Hz, FD = 0.77 (±0.05), |9034
% change using ABF
Chapter 2
100
50
**
** **
**
0
-50
-100
**
*
EC
EOF
RMS
MV
F95%
FD
|90-Mdir|
ECF
Figure 4 - Effects of ABF on COP parameters in all three sensory conditions tested. Boxplots describe the
distribution of the percent changes of the five parameters across the population. Small circles indicate
outlying values. *: p < 0.05; **: p < 0.01.
Mdir| = 17.0 (±15.9) deg; and with ABF: RMS = 12.5 (±3.6) mm, MV = 28.1 (±7.6) mm/s, F95%
= 1.73 (±0.15) Hz, FD = 0.75 (±0.03), and |90-Mdirj| = 19.0 (±14.9) deg.
35
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
Discussion
We have developed and preliminarily tested an ABF system that sends trunk acceleration
information to users to help them correct postural sway during stance. This acoustic information
helped subjects reduce postural sway, especially when visual and sensory information were
compromised by eye closure and stance on foam. The instrument met requirements for an
adequate biofeedback system: adequate bandwidth and sensitivity, convenient feedback
signal generation, and lightweight portability. None of the subjects had problems learning
how to use the ABF system, and the 1-min practice trials were adequate to teach them how
to use ABF to reduce their sway while quietly standing. The efficacy of ABF appears to depend
on the availability of alternate sensory information since the more subjects were unstable in a
sensory condition, the more that they improved their balance with ABF. This finding suggests
that subjects use ABF to partially substitute for the lack of visual information and/or for the
unreliability of somatosensory information while they try to maintain postural control.
The results reported here were from experiments with normal, healthy subjects who have
extensive sensory and functional redundancy in their postural system. We hypothesized that
our ABF device would help subjects with sensory deficits improve postural sway even more,
and subsequent studies of ABF experiments with bilateral vestibular loss subjects confirmed
this hypothesis [13]. In the present study the improvements in stance were probably due
to a change in postural control strategies because sway variables measured with ABF were
consistent with smaller (see decrease in RMS) and more frequent (see increase in MV and
F95%) postural corrections [14]. In accord with this result is also the decrease in correlation
between COP and acceleration signals observed in the ECF condition. This decrease may
reflect a moderate decline in the simple ankle strategy to maintain balance [15] in factor of
more complex control; experiments aimed at investigating this hypothesis are in progress.
However, it is possible that the attention of these subjects (which was not measured in the
protocol reported here) may also have contributed, at least in part, to their improved balance
while using ABF.
Many earlier biofeedback systems used audio alarms to notify the user of abnormal
values of monitored parameters (e.g. [16]). The present ABF system is novel in the use of
nonlinear coding functions and in the customization of these functions to each subject and
task. Preliminary results suggest this ABF device may be a useful tool for rehabilitation in
the clinic, home-care setting, and community during mobility training. The use of ABF may
36
Chapter 2
become attractive for rehabilitation, especially if it is found to favor neural plasticity in motor
control [2]. In other words, a person with impaired abilities to control posture could practice
with ABF to achieve better postural control when not using ABF.
Plans are underway to improve the current ABF system by making it wireless for
increased portability and for enabling remote control and remote monitoring. Different
sonification procedures will also be tested in the near future. In particular, 3-D generated sound
with a headrelated transfer function or immersive sound will be investigated. In addition,
since the current ABF system may interfere with hearing for communication purposes or
may be unsuitable for people with hearing deficits. Other sound-delivery processes will be
investigated, including bone mastoid vibration.
The strong correlation between COP and acceleration signals suggests that the sensory
unit could be developed for use as a portable, miniaturized force plate [17], which may be
helpful for remote monitoring such as for elderly persons and persons with postural and
mobility disorders.
37
Audio-Biofeedback for Balance Improvements: an Accelerometry-Based System
Bibliography
[1]
P. G. Morasso, L. Baratto, R. Capra, and G. Spada, “Internal models in the control of posture,” Neural Netw,
vol. 12, pp. 1173–1180, Oct. 1999.
[2]
A. Shumway-Cook, D. Anson, and S. Haller, “Postural sway biofeedback: its effect on re-establishing stance
stability in hemiplegic patients,” Arch Phys Med Rehabil, vol. 69, no. 6, pp. 395–400, Jun. 1988.
[3]
C. Wall, III and M. S. Weinberg, “Balance prostheses for postural control,” IEEE Eng Med Biol, vol. 22, pp. 84–90,
Mar./Apr. 2003.
[4]
P. Bonato, “Wearable sensors/systems and their impact on biomedical engineering,” IEEE Eng Med Biol Mag,
vol. 22, pp. 18–20, May/Jun. 2003.
[5]
L. Bianchi, F. Babiloni, F. Cincotti, M. Arrivas, P. Bollero, and M. G. Marciani, “Developing wearable bio-feedback
systems: a general-purpose platform,” IEEE Trans Neural Syst Rehabil Eng, vol. 11, no. 2, pp. 607–609, Jun.
2003.
[6]
M. S. Schwartz and R. P. Olson, “A historical perspective on the field of biofeedback and applied
psychophysiology,” in Biofeedback—A Practitioner’s Guide, M. S. Schwartz and F. Andrasik, Eds. New York:
The Guilford Press, 2003, pp. 3–19.
[7]
C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical
sensors using vibrotactile feedback of tilt,” IEEE Trans Biomed Eng, vol. 48, no. 10, pp. 1153–1161, Oct.
2001.
[8]
D. Giansanti, V. Macellari, and G. Maccioni, “The development and test of a device for the reconstruction
of 3-D position and orientation by means of a kinematic sensor assembly with rate gyroscopes and
accelerometers,” IEEE Trans Biomed Eng, vol. 52, no. 7, pp. 1271–1277, Jul. 2005.
[9]
R. E. Mayagoitia, J. C. Lotters, P. H. Veltink, and H. Hermens, “Standing balance evaluation using a triaxial
accelerometer,” Gait Posture, vol. 16, pp. 55–59, 2002.
[10]
D. Deutsch, Ear and Brain: How We Make Sense of Sounds. NewYork: Springer-Verlag, 2003.
[11]
L. Rocchi, L. Chiari, and A. Cappello, “Feature selection of stabilometric parameters based on principal
component analysis,” Med, Biol Eng, Comput, vol. 42, no. 1, pp. 71–79, 2004.
[12] E. V. Gurfinkel, “Physical foundations of stabilography,” Agressologie, vol. 14, pp. 9–14, 1973.
[13] M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback improves balance in patients with bilateral vestibular
loss,” Arch Phys Med Rehabil, vol. 86, no. 7, pp. 1401–1403, Jul. 2005.
[14] M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A. Cappello, “Influence of a portable audio-biofeedback
device on structural properties of postural sway,” J Neuroeng Rehabil, vol. 2, 2005.
[15]
F. B. Horak and L. M. Nashner, “Central programming of postural movements: adaptation to altered supportsurface configurations,” J Neurophysiol, vol. 55, no. 6, pp. 1369–1381, 1986.
[16] M. S. Wong, A. F. Mak, K. D. Luk, J. H. Evans, and B. Brown, “Effectiveness of audio-biofeedback in postural
training for adolescent idiopathic scoliosis patients,” Prosthet Orthot Int, vol. 25, no. 1, pp. 60–70, 2001.
[17] R. Moe-Nilssen and J. L. Helbostad, “Trunk accelerometry as a measure of balance control during quiet
standing,” Gait Posture, vol. 16, pp. 60–68, 2002.
38
Chapter 2
39
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
40
Chapter
Chapter 3
Influence of a Portable AudioBiofeedback Device on Structural
Properties of Postural Sway
Most of the content of this chapter has been published in: M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A.
Cappello, “Influence of a portable audio-biofeedback device on structural properties of postural sway,” J Neuroeng
Rehabil, vol. 2, 2005.
41
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
42
Chapter 3
Abstract
Good balance depends on accurate and adequate information from the senses. When
sensory information is limited or unreliable balance may become critical. One way to substitute
missing sensory information for balance is with biofeedback systems. We previously reported
that audio-biofeedback (ABF) has beneficial effects in subjects with profound vestibular loss,
since it significantly reduces body sway in quiet standing tasks
In this paper, we present the effects of a portable prototype of an ABF system on healthy
subjects’ upright stance postural stability, in conditions of limited and unreliable sensory
information. Stabilogram diffusion analysis, combined with traditional center of pressure
analysis and surface electromyography, were applied to the analysis of quiet standing tasks
over a Temper foam surface.
These analyses provided new evidence that ABF may be used to treat postural instability.
In fact, the results of the stabilogram diffusion analysis suggest that ABF increased the amount
of feedback control exerted by the brain for maintaining balance. Interestingly, the resulting
increase in postural stability was not at the expense of leg muscular activity, which remained
almost unchanged.
Examination of the stabilogram diffusion analysis and the EMG activity supported the
hypothesis that ABF does not induce an increased stiffness (and hence more co-activation) in
leg muscles, but rather helps the brain to actively change to a more feedback-based control
activity over standing posture.
43
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
Introduction
Maintaining balance is a complex task accomplished by the brain through the fusion
and interpretation of sensory information. When sensory information from vestibular,
somatosensory, and visual systems [1-3] are not accurate and/or adequate, balance will be
compromised. Although, in many cases, the loss of peripheral sensory information is not
curable or reversible, the brain can compensate for the loss of sensory information by relying
more on the other sensory channels [4;5].
The purpose of biofeedback (BF) systems for postural control is to provide additional
sensory information about body equilibrium to the brain [6]. In the last few years, different
sensors, encoding algorithms, and information restitution devices have been combined to
develop promising BF systems for postural control [7-9]. The major design goals were focused
on portability, usability, economy, and effectiveness in balance improvements [8;10-12].
The development of these BF systems has been facilitated by the availability of
lightweight, miniaturized, and economical sensors such as accelerometers, inclinometers,
and gyroscopes [13]. The use of these sensors makes BF devices inexpensive, unsusceptible
to shadowing effect, and not limited in the measurement field, in contrast to dynamometric
platforms and motion analysis systems, which are commonly used in laboratory settings
[14;15]. In addition, due to their size and weight, these sensors can measure body segment
movement without hindering natural motor execution.
More detail is needed for understanding how biofeedback information interacts with
the brain or, from a neuroscience perspective, how the brain uses artificial BF information and
combines it with natural sensory information. We believe that understanding this interaction
is fundamental for further developing effective BF systems.
An interesting analysis in the understanding of how the brain may use BF information for
postural control was proposed by Collins and De Luca in 1993 [16]. These authors developed
a statistical-biomechanics method for analyzing force platform data recorded during quiet
standing, called stabilogram diffusion analysis (SDA). SDA was applied to center of pressure
(COP) data and it disclosed that COP tends to drift away from a relative equilibrium point over
short-term observation intervals (less than 1-second long), whereas COP tends to return to a
relative equilibrium point over long-term observation intervals. These results took Collins and
De Luca to suggest that the motion of the COP is not purely random, and that SDA may be
44
Chapter 3
able to give insight on the amount of open-loop and closed-loop postural control applied by
the central nervous system for maintaining balance [17]. SDA was used several contexts, e.g.
to evaluate the effect of spaceflight [18], visual input [19;20], and age-related changes [21;22]
on postural stability. In 2000, Chiari developed and validated a new nonlinear model for
extracting parameters from SDA diagrams, reducing from 6 to 2 the number of the parameters
used to characterize the structural properties of COP [20]. In 2004, Rocchi found that these
new parameters may be useful adjuncts to evaluate postural control strategies in patients
with Parkinson’s disease and may allow the comparison of different deep brain stimulation
electrode sites based on their effect on structural properties of the COP [23].
In this paper, we investigate the effect on postural stability of a portable, accelerometrybased, audio biofeedback (ABF) system recently developed by the authors [9]. Standing with
eyes closed on TemperTM foam will be used to evaluate the effects of artificial auditory cues to
enhance the reduced (from the eyes) and masked (from the feet) natural sensory information.
Measurements include COP recorded by a force platform under the feet, trunk acceleration
measured by the ABF sensors, and EMG signals from the leg muscles. SDA according to Chiari
et al. [20], traditional COP analysis [24], and muscle activation analysis according to Olney &
Winter [25] were performed in order to evaluate the effect of ABF on healthy young subject’s
upright posture.
These analyses were aimed to answer two questions: (1) do structural properties of
postural sway change with ABF? And, if so, (2) in which way will this help in understanding the
mechanisms underlying ABF efficacy and in improving the design of a rehabilitation strategy
for balance disorders? In this paper, we present evidence that supports the hypothesis that
ABF does not simply induce a purely biomechanical increase in stiffness (and hence more
co-activation) in the leg muscles, but rather ABF helps the brain actively adapt its control
activity over standing posture.
45
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
Methods
Participants
Eight healthy subjects participated to this experiment (5 males and 3 females, aged
23.5±3.0 yrs, range 21-28 yrs). All participants were free from any neurological, orthopedic,
hearing, or vestibular disorder. Informed consent form was obtained from each subject.
The form was prepared in accordance with the Oregon Health & Science University Ethical
Committee and respected the declaration of Helsinky, 1964.
Apparatus and procedure
Subjects performed 10, 60-second trials standing with eyes closed on TemperTM, 4’’-thick
foam. COP displacement was recorded with an AMTI OR6-6 force plate. An ABF system [9] was
used to provide subjects with additional balance information related to trunk acceleration. The
ABF system used a sensor, based on 2-D accelerometers (Analog Device ADXL203) mounted
on the subject’s back (L5), to create an audio stereo sound representing the acceleration
sensed along the anterior-posterior (AP) and the medial-lateral (ML) direction. A laptop,
Toshiba Celeron 2.3 GHz, was dedicated to convert the accelerations into stereo sounds.
Commercial headphones were used by the subjects to listen to the ABF sound. The ABF system
is described in detail in [9] and illustrated in Figure 1. In short, the stereo sound provided
by the ABF system consisted of two sine waves, one for the left ear channel and one for the
right ear channel. Pitch, volume and left/right balance of the stereo sound were modulated
to represent the 2-D acceleration information. Specifically, when the subject swayed forward,
and consequently the acceleration increased in the anterior direction, the sound got louder
in volume and higher in pitch. When the subject swayed backward, and consequently the
acceleration increased in the posterior direction, the sound got louder in volume and lower
in pitch. When the subject moved right and, consequently, the acceleration increased in the
right direction, the sound got louder in the right ear channel and lower in the left one. When
the subject moved left, and consequently the acceleration increased in the left direction, the
sound got louder in the left ear channel and lower in the right one. The sound dynamics was
optimized for each trial by taking as a reference the first 10-second recordings of each trial.
The equations used for the pitch, volume, and left/right balance modulation can be found
in [9]. Each subject was instructed to maintain balance during the trials by taking advantage
of the ABF information, when available. Five trials with ABF and 5 trials without ABF were
46
Chapter 3
performed in randomized order by each subject. Before
the experimental session, the subjects were instructed
on how ABF codes trunk acceleration into sound, and
performed free-movement trials until they felt confident
in performing the full experiment.
Data recording
For each standing trial, ground reaction forces and
torques were recorded from the force place with a 100Hz sample frequency. COP displacement was processed
offline from the force plate data after applying a 10-Hz
cut-off, low-pass Butterworth filter. Accelerations
along AP and ML direction were collected with a 100
Hz sample frequency after applying a low-pass filter
with a 20-Hz cut-off. EMG was recorded from right leg
muscles, tibialis (TI), soleus (SO), and gastrocnemius
(GA) with two surface electrodes fixed about 2 cm
apart along the length of each muscle belly; the
ground electrode was fixed on a bony area of the
right hallux. The EMG signals were amplified 20000
times, band-pass filtered (71-2652 Hz), integrated
and full-wave rectified with a 6th order Butterworth
low pass with a cut-off of 100Hz.
Figure 1 – ABF system device and protocol
The ABF consisted of (1) a sensor mounted
on the trunk and sensing acceleration along
AP and ML axes, (2) a laptop acquiring
acceleration from the sensor and processing
the ABF sound, (3) a pair of earphones the
subject wears for listening to the sound. In
this figure is also shown the protocol where a
healthy normal subject is standing on a foam
placed on a force plate with eyes closed. Finally,
in the figure bottom right statokinesigrams
in condition with and without ABF from a
representative subject are shown.
Data analysis
From AP COP data, the root mean square distance (COP-RMS) and the frequency
comprising the 95% of the power (F95%) were extracted according to Prieto et al. [24].
From the acceleration sensed at trunk level along AP direction we computed the root
mean square value (Acc-RMS).
In addition, two stochastic parameters were included in the analyses. These parameters
characterize a previously developed model that describes with continuity the transition
among the different scaling regimes found in the COP time series [26]. The model is described
by the following equation:
V(Δt) = K Δt2H(Δt)
where V(Δt) is the variance of COP displacement, computed at time-lag Δt, and H is the
scaling exponent, also called Hurst exponent. This is assumed to follow a sigmoid law in the
time interval (Δt):
47
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
H(Δt)=
log 2
log [ 2 (1+Δt/ΔTc) ]
In this way, the features extracted from COP data are the following (see [20] for more
details):
- K is an estimate of the diffusion coefficient of the random process obtained by sampling
the COP time series at the sampling frequency 1/ΔTc.
- ΔTc represents the time-lag at which the real process corresponds to a purely random
behavior, and where it switches from a persistent (positively correlated, and hence interpreted
in terms of feed-forward control) to an anti-persistent (negatively correlated, and hence
interpreted in terms of feedback control) behavior [16].
Mean muscular activity was calculated from the full wave rectified EMG of each muscle.
For each subject and each muscle, muscle activity was expressed in percentage in reference
to the trial with maximal activity recorded. This made possible the comparison of muscle
activity among the different subjects. The EMG signals were further processed applying a
low pass-filter with a 2 Hz cut-off in order to obtain tension curves according to [25]. These
tension curves were cross-correlated to determine the amount of co-activation between the
muscles recorded.
Statistical analysis
Paired T-tests were performed to determine the effect of ABF on the different parameters
extracted from COP, acceleration and EMG data collected. The threshold for statistical
significance was set to p=0.05.
48
Chapter 3
Results
ABF effect on sway parameters
Subjects’ confidence and comfort
All participants reported ABF sound was
comfortable and its way of representing the
information was intuitive. In fact, none of the subject
needed more than two, free-movement trials before
feeling ready to start the experiment.
Subjects’ sway
A B F s i g n i f i c a n t l y i n f l u e n ce d s u b j e c t s’
performance on the foam. The percentage change
induced by ABF on all sway parameters, either
measured at the trunk level with the accelerometer
or at the feet level with the force platform, is shown
in Figure 2. Figure 2 also reports significance levels
of the parameter changes occurred while using
the ABF. The general results shown in Figure 2 are
specified in detail in the following.
% difference using ABF
10
0
COP-RMS
Acc-RMS
K
F95%
-10
-20
-30
-40
*
*
**
COP parameters
Acceleration parameter
SDA parameters
*
Figure 2 - Effect of ABF on sway
The effect of using ABF on the sway
parameters is reported in percentage. COPRMS and F95% were extracted from the AP
COP displacement according to [24]. AccRMS was extracted from AP acceleration
recorded at trunk level (L5). K and Δtc were
derived by applying the method proposed
by [20] on the SDA diagrams [16]. Asterisks
indicate statistical significance: * p<0.05 and
** p<0.01. The reductions of K, COP-RMS and
Acc-RMS are a consistent evidence of the
reduction of sway amplitude shown by the
subject using ABF. The increasing of F95%
suggests that the postural control applied by
the CNS when ABF is available was increased.
The reduction of ΔTc suggests a major active
closed-loop postural control exercised by the
CNS.
Center of Pressure analysis
Center of pressure displacement in the AP direction was significantly influenced by ABF.
T-tests results revealed significant effects of ABF on COP-RMS (p=0.015). This effect is shown
by a consistent reduction of COP-RMS for 7 out of 8 subjects as shown in Table 1 (column 7).
Average reduction of COP-RMS was 10.7%. Columns 1 and 4 of Table 1 also show the subjectby-subject values of COP-RMS without and with ABF, respectively. The last three subjects (#6,
#7, #8) were females and showed smaller COP-RMS, as expected considering their smaller
heights [26].
F95% increased with ABF for 7 out of 8 subjects ( Table 1, column 8) but this result
was not significant (p=0.42). The values of F95% are also reported for each subject in both
conditions (Table 1, columns 2 and 5). Average increase of F95% due to ABF was 6.2% as
shown in Figure 2.
49
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
Table 1 – ABF effect on sway parameters. Standard deviations are indicated in parenthesis.
COP-RMS
(NO - ABF) [mm]
F95 %
(NO - ABF)
[Hz]
Acc-RMS
(NO - ABF)
[mm/s2]
COP-RMS (ABF)
[mm]
F95 %
(ABF)
[Hz]
Acc-RMS
(ABF)
[mm /s2]
% COP-RMS
difference
% F95 %
difference
% Acc-RMS
difference
Subj. #1
10.79 (2.84)
0.99 (0.05)
137 (48)
9.57 (1.86)
1.18 (0.16)
118 (13)
-11.2
19.1
-14.1
Subj. #2
9.91 (2.77)
1.20 (0.29)
142 (27)
9.50 (2.26)
1.30 (0.20)
120 (23)
-4.1
8.7
-15.6
Subj. #3
9.21 (2.94)
1.16 (0.14)
121 (23)
8.61 (1.42)
1.37 (0.07)
113 (21)
-6.5
18.0
-7.0
Subj. #4
10.23 (1.50)
1.43 (0.08)
117 (30)
8.80 (1.74)
1.49 (0.12)
100 (12)
-13.9
4.1
-14.6
Subj. #5
8.50 (0.93)
1.49 (0.22)
143 (46)
6.90 (1.35)
1.53 (0.28)
115 (19)
-18.8
2.6
-19.3
Subj. #6
9.62 (1.55)
1.34 (0.30)
126 (43)
7.35 (0.88)
1.34 (0.09)
89 (20)
-23.6
0.0
-29.2
Subj. #7
6.37 (1.48)
1.60 (0.07)
64 (8.3)
5.19 (0.59)
1.94 (0.12)
51 (4.7)
-18.5
20.8
-20.1
Subj. #8
6.08 (1.19)
1.78 (0.25)
48 (6.3)
6.75 (1.41)
1.37 (0.16)
39 (3.8)
10.9
-23.1
-17.3
Average
8.84 (1.75)
1.37 (0.26)
112 (36)
7.83 (1.54)
1.44 (0.15)
93 (31)
-10.7 (10.9)
6.2 (14.4)
-17.2(6.3)
Acceleration analysis
Acceleration sensed at trunk level (L5) in AP
direction was significantly reduced by ABF. T-test
results also revealed significant effects of ABF on
Acc-RMS (p=0.0009). Acc-RMS was reduced by ABF
across all subjects, as shown in Table 1 (last column).
Average reduction of Acc-RMS was 17.2% (Figure 2).
Columns 3 and 7 of Table 1 also show the subjectby-subject values of Acc-RMS without and with ABF,
respectively. The last three subjects were females and
showed smaller Acc-RMS, as expected considering
their smaller heights [26].
50
30
20
#7
#1
#3
10
#5
#2
#6
0 COP-RMS - % change using ABF
-20
-10
F95% - % change using ABF
#4
-30
10
20
-10
-20
#8
-30
Figure 3 – Antithetic behavior of subject #8.
On the horizontal axis COP-RMS percentage
change using ABF is reported whereas on
the vertical axis F95% percentage chance
using ABF is reported. The values of each
subject from Table 1 are plotted. Subject #8
behaves antithetically to the other subjects.
102
K
Variance [mm2]
It is worth noting that subject #8 behaved as
an outlier (Figure 3), compared to the other subjects
since she was the only one who showed opposite
changes in COP-RMS and F95% while using ABF.
Performing the T-Tests with this outlier eliminated
increased the effect of ABF on COP-RMS (p=0.002),
and on F95% (p=0.02). These results better match
the results already published in [9]. The outlying
behaviour of subject #8 will be investigated further
in the discussion.
101
SD diagram
Parameterization
100
SD diagram
Parameterization
Tc
10-1
10-2
10-1
100
101
Tc [s]
Figure 4 – Effect of ABF on open-loop and
closed-loop control.
SDA diagrams for one representative subject.
Two conditions are reported: without ABF
(black) and with ABF (gray). The behavior of the
parameters K and ΔTc used to parameterize
the SDA diagrams is also shown. This figure
suggests that, using ABF, subjects decrease the
amount of sway by increasing the closed-loop
(feedback) posture control.
Chapter 3
Stabilogram diffusion analysis
A. Muscle activity
Muscle activity analysis
Percent activity
ΔTc characterizing the SDA diagram,
were both significantly decreased by
ABF (Figure 2). Average K reduction
was 9.3% (p=0.02), whereas average
ΔTc reduction was 33.9% (p=0.018).
Table 2 reports the subject-by-subject
values of K and ΔTc in both conditions
tested. Subject #8 and subject #7 are
the only ones who showed a slight
increase in K.
90
p=0.79
p=0.17
ABF
80
70
60
50
40
30
20
10
0
TI
Coefficient of determination r2
SDA diagrams plotted from
AP COP data, were also significantly
influenced by ABF (Figure 4). As a
consequence, the parameters K and
NO ABF
p=0.64
100
GA
SO
B. Muscle co-activation
0.7
NO ABF
p=0.42
ABF
0.6
0.5
0.4
0.3
p=0.63
p=0.51
TI-GA
TI-SO
0.2
0.1
0.0
GA-SO
Figure 5 – Effect of ABF on muscle.
Estimates of muscular co-activation (Fig. 5A) for different pair
of muscles (TI-GA, TI-SO, GA-SO) and muscle activity (Fig. 5B)
are shown. Average values are reported for trials with (light
gray) and without (dark gray) ABF. Error bars represent standard
deviation. As shown in Figure 5A, using ABF does not change
significantly the co-activation between the muscles analyzed (p
values from T-Test are reported). This suggests that the major
amount of postural corrections induced by ABF does not involve
a major co-activation of the muscles TI, GA, and SO in the leg.
As shown in Figure 5B, using ABF does not change significantly
the activity of the muscles analyzed (p values from T-Test are
reported). This suggests that the major amount of postural
corrections induced by ABF does not involve a major average
activity of the muscles TI, GA, and SO in the leg.
Muscle activity of TI, GA, and SO
was not influenced by ABF. Overall,
the mean activity, expressed as a
percentage of the maximal activity
recorded from each single muscle
across all the trials of a subject, did not
change significantly due to ABF (see
Figure 5A). TI activity showed a trend
toward increasing in trials with ABF (p=0.17) but this change was particularly clear only for
subjects #4 and #7.
Table 2 – ABF effect on SDA parameters. Standard deviations are indicated in between parenthesis.
K (NO-ABF)
[mm2]
Δtc (NO-ABF)
[s]
K (ABF)
[mm2]
Δtc (ABF)
[s]
%K
difference
% Δtc
difference
Subj. #1
100 (57)
0.42 (0.21)
86 (15)
0.38 (0.17)
-14.6
-9.9
Subj. #2
70 (29)
0.51 (0.31)
66 (20)
0.41 (0.34)
-7.4
-20.5
Subj. #3
75 (41)
0.52 (0.29)
65 (20)
0.29 (0.12)
-13.3
-45.3
Subj. #4
80 (21)
0.81 (0.46)
70 (14)
0.39 (0.14)
-11.1
-52.0
Subj. #5
47 (13)
0.32 (0.08)
39 (10)
0.26 (0.16)
-18.1
-19.7
Subj. #6
64 (12)
0.27 (0.08)
61 (9)
0.20 (0.09)
-5.7
-26.1
Subj. #7
32 (7)
0.17 (0.06)
34 (9)
0.09 (0.01)
6.6
-47.4
Subj. #8
35 (14)
0.29 (0.09)
38 (13)
0.19 (0.06)
5.8
-34.3
Average
63 (23)
0.41 (0.20)
57 (18.5)
0.27 (0.11)
-9.3 (9.2)
-33.9 (15.3)
51
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
Muscle co-activation of ankle agonists-antagonists did not change significantly due to
the ABF (see Figure 5B). Co-activation between TI and GA was small both with (r2=0.11) and
without (r2=0.08) ABF. Similarly small was the co-activation between TI and SO with (r2=0.14)
and without (r2=0.09) ABF. As expected, co-activation between GA and SO was instead large
(r2=0.39 in trials with ABF and r2=0.46 in trials without ABF). Figure 5B reports the coefficient
of determination r2, which indicates the amount of muscular co-activation, for all pairs of
muscles analyzed in trials with and without ABF.
A. Muscle activity
p=0.08
100
*
p=0.049
90
Percent activity
NO ABF
p=0.51
ABF
80
70
60
50
40
30
20
10
0
Coefficient of determination r2
TI
52
GA
SO
B. Muscle co-activation
0.7
NO ABF
0.6
0.5
p=0.20
p=0.23
0.4
p=0.36
0.3
0.2
0.1
0.0
TI-GA
TI-SO
GA-SO
ABF
Figure 6 – Muscle activity and co-activation in subject
#8.
The antithetic behavior of subject #8 for muscles coactivation, (Fig.6A), and for muscles activity, (Fig. 6B) is
shown. Figure 6a reports the estimates of muscular coactivation for different pair of muscles: TI-GA, TI-SO, and
GA-SO. Average values are reported for trials with (light
gray) and without (dark gray) ABF. Error bars represent
standard deviation. Even if co-activation looks higher
in trials with ABF for all couples of muscles while using
ABF, muscles co-activation does not change significantly
(p values from T-Test are reported; since the number of
samples is five it is convenient to report also the powers
which were respectively: 0.20, 0.14, 0.23). This suggests
that a major amount of co-activation of the muscles
TI, GA, and SO was exercised by this subject while using
ABF. Figure 6B reports the estimates of muscular activity
for TI, GA, and SO muscle. Average values expressed in
percentage are reported for trials with (light gray) and
without (dark gray) ABF. Error bars represent standard
deviation. The percent activity was calculated taking as
one-hundred-percent reference the trial with the highest
muscle activation recorded. Even if muscles activity looks
higher in trials with ABF for all muscles, only SO activity
changed significantly while using ABF (p values from
T-Test are reported; since the number of samples is five,
it is convenient to report also the powers which were
respectively: 0.09, 0.41, 0.53). This suggests that a major
amount of activity of the muscles TI, GA, and SO was
exercised by this subject while using ABF.
Chapter 3
Discussion
Using the proposed ABF device, all healthy subjects included in this study could sway
less when standing in a particularly challenging condition, with vision unavailable and
somatosensation partly unreliable. All subjects, in fact, reduced their AP Acc-RMS (see Table
1). In this way, subjects were further from their stability limits and, consequently, more stable.
Trunk stabilization also entailed the need of smaller corrective torques at the ankles, and
hence smaller COP displacements. This is proved by the fact that all subjects but one (Subj.
#8) showed a significant decrease in AP COP-RMS (Fig. 2). During ABF, postural corrections in
leg muscles were likely smaller but more frequent in number, as suggested by the increase
in F95% of the COP, even if the EMG signals available did not clearly confirm this possibility.
Future studies involving more sophisticated techniques for the acquisition and analysis of
the EMG signals will be needed to validate this hypothesis. This result suggests that ABF may
partially substitute for the lack of visual and somatosensory information for postural control
by taking the postural control system towards a new steady state associated with a different
control strategy.
Examination of the SDA and the EMG activity supported the hypothesis that ABF does
not simply induce an increased stiffness (and hence more co-activation) in leg muscles, but
rather helps the brain to actively change to a more feedback-based control activity over
standing posture. Representative SDA diagrams reported in Figure 4 suggest that ABF
contributes to a general reduction of both the diffusion coefficient K and the transition time
ΔTc. Downward shifts of the SDA diagrams, described by smaller diffusion coefficients, reflect
a reduced stochastic activity of the COP, and hence a more tightly regulated control system
[16]. Shorter transition times reflect an earlier switching between persistent and antipersistent
behaviors, and hence more prompt reactions to perturbations of the postural control system
[27]. In summary, these results disclose, as a consequence of ABF: 1) an increase in stability,
and 2) a more prominent role for feedback control over feed-forward control. Hence, the
solution proposed by the brain after ABF seems to involve more feedback control for a more
stable sway.
Interestingly, this result is partly different from the one observed by Rougier in quiet
stance experiments with visual BF [28]. In that condition, with BF, SDA diagrams only changed
some local properties (local slopes) over short or long observation intervals but did not shift
significantly, meaning that one may expect that K is not changing that much. Further, closedloop control operated over longer observation-times, suggesting that feed-forward control is
53
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
expanding over feedback control. Such a different behavior may find an explanation in the
peculiar role, not just a simple redundancy, of different senses in multi-sensory integration for
the control of posture [29]. Whereas vision induces alertness of the outer environment and
hence pushes towards predictions of forthcoming events in the scene (feed-forward control)
[30]. In contrast, hearing, compared to vision, may be more important for postural reactions to
disturbing stimuli (feedback control). This result can also be related to the different processing
times required by the central nervous system for visual and auditory stimuli with auditory
reaction times significantly faster than visual reaction times. Finally, another factor which may
explain the different outcomes of the two BF-studies is the selection of two, different, input
variables (COP for visual BF and Acceleration from the trunk for ABF). It is widely accepted
that upper- and lower- body segments are controlled separately [31].
Both predictive (feed-forward) and reactive (feedback) control need to be used in order
to have an adequate interaction with the environment. For this reason, it’s hard to tell if
ABF is preferred to visual BF, or vice versa. Rather, the point is that it could be important, in
a rehabilitation setting, to identify which one of the two components of postural control
need more reinforcement or substitution in a particular patient, and consequently design
an optimized BF treatment.
The outlying results observed for Subj. #8 need to be discussed individually. This woman
in fact did not decrease COP-RMS and K, and did not increase F95%, even if, similarly to
the other subjects, she decreased Acc-RMS and ΔTc (these changes were consistent across
the whole population). Hence, with ABF she actually swayed less and she showed the same
increase of feedback control. Nonetheless, either due to her small body size or to a slightly
different control scheme, she obtained these goals with a different strategy. Figure 6 reports
her muscle activities and co-activations. It can be seen how she generally improves muscle
activity with ABF (Figure 6A), in particular with a large increase in the activity of posterior
muscles, GA and SO. It should be noted, however, that also the estimated co-activations (Figure
6B) look pretty dissimilar compared with the ones of the other subjects, shown in Figure 5B.
Particularly low is the co-activation of agonists muscles GA-SO without ABF, which ABF partly
contributes to enlarge. For all these reasons her postural behavior in the proposed task should
be looked as an outlying behavior and more analyses are needed, on a larger population, to
assess the real influence of body size or usual control setting on the responsiveness to ABF.
Many earlier biofeedback systems used audio alarms to notify the user of abnormal
values of monitored parameters (e.g. [32]). The present ABF system is novel in the use of
nonlinear coding functions and in the customization of these functions for each subject and
task. Although the current ABF system may interfere with use of hearing for communication,
it may be quite useful during the rehabilitation and training process. Plans are underway to
improve the current ABF system by making it wireless for increased portability and equipping
it with a communication module for remote control, recording, and monitoring. Different
54
Chapter 3
sonification procedures will also be tested and compared in a near future. Specifically, 3-D
generated sound with a HRTF (Head Related Transfer Function) or immersive sound may be
even more effective signal for improving stance balance.
In conclusion, we have investigated the attributes of a portable instrument that feeds
back trunk acceleration to help subjects reducing their postural sway during stance. The
instrument meets requirements for an adequate biofeedback system that may find interesting
applications not only as a rehabilitation device in the clinic, but also in the home care setting,
and when doing community mobility training outside the traditional clinic setting. In fact,
it has appropriate bandwidth and sensitivity, smoothness and delay of the acoustic signal
generator, and portability. Acoustic information related to trunk movement allowed subjects
in the present experiment to increase postural stability when sensory information from
vision and the surface were compromised by eye closure and stance on foam. We provided
evidence that the balance improvement was not simple stiffening at the ankle, but rather the
brain actively adapted its control activity over standing posture with more feedback-based
control.
55
Influence of a Portable Audio-Biofeedback Device on Structural Properties of Postural Sway
Bibliography
[1]
F. B. Horak, C. L. Shupert, and A. Mirka, “Components of postural dyscontrol in the elderly: a review,” Neurobiol
Aging, vol. 10, no. 6, pp. 727-738, Nov.1989.
[2]
S. R. Lord and J. A. Ward, “Age-associated differences in sensori-motor function and balance in community
dwelling women,” Age Ageing, vol. 23, no. 6, pp. 452-460, Nov.1994.
[3]
H. Corriveau, F. Prince, R. Hebert, M. Raiche, D. Tessier, P. Maheux, and J. L. Ardilouze, “Evaluation of postural
stability in elderly with diabetic neuropathy,” Diabetes Care, vol. 23, no. 8, pp. 1187-1191, Aug.2000.
[4]
F. B. Horak and Macpherson J.M., “Postural equilibrium and orientation,” in Handbook of Physiology. Rowell
RB and Shepherd J.T., Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[5]
F. B. Horak and F. Hlavacka, “Somatosensory loss increases vestibulospinal sensitivity,” J Neurophysiol, vol.
86, no. 2, pp. 575-585, Aug.2001.
[6]
S Moore and M. Woollacott, “The use of biofeedback to improve postural stability,” Phys Ther Practice, vol.
2, pp. 1-19, 1993.
[7]
C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical sensors
using vibrotactile feedback of tilt,” IEEE Trans Biomed Eng, vol. 48, no. 10, pp. 1153-1161, Oct.2001.
[8]
M. Tyler, Y. Danilov, and Bach-y-Rita P, “Closing an open-loop control system: vestibular substitution through
the tongue,” J Integr Neurosci, vol. 2, no. 2, pp. 159-164, Dec.2003.
[9]
L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[10] M. Bazzarelli, N. Durdle, E. Lou, J. Raso, and D. Hill, “A wearable networked embedded system for the
treatment of scoliosis,” Stud Health Technol Inform., vol. 91, pp. 383-386, 2002.
[11] P. P. Kadkade, B. J. Benda, P. B. Schmidt, and C. Wall, III, “Vibrotactile display coding for a balance prosthesis,”
IEEE Trans Neural Syst Rehabil Eng, vol. 11, no. 4, pp. 392-399, Dec.2003.
[12] M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback improves balance in patients with bilateral vestibular
loss,” Arch. Phys. Med. Rehabil., vol. 86, no. 7, pp. 1401-1403, July2005.
[13] H. J. Luinge and P. H. Veltink, “Inclination measurement of human movement using a 3-D accelerometer
with autocalibration,” IEEE Trans Neural Syst Rehabil Eng, vol. 12, no. 1, pp. 112-121, Mar.2004.
[14] R. E. Mayagoitia, J. C. Lotters, P. H. Veltink, and H. Hermens, “Standing balance evaluation using a triaxial
accelerometer,” Gait Posture, vol. 16, no. 1, pp. 55-59, Aug.2002.
[15] R. E. Mayagoitia, A. V. Nene, and P. H. Veltink, “Accelerometer and rate gyroscope measurement of kinematics:
an inexpensive alternative to optical motion analysis systems,” J Biomech, vol. 35, no. 4, pp. 537-542,
Apr.2002.
[16] J. J. Collins and C. J. De Luca, “Open-loop and closed-loop control of posture: a random-walk analysis of
center-of-pressure trajectories,” Exp Brain Res, vol. 95, no. 2, pp. 308-318, 1993.
[17] J. J. Collins and C. J. De Luca, “Upright, correlated random walks: A statistical-biomechanics approach to
the human postural control system,” Chaos, vol. 5, no. 1, pp. 57-63, Mar.1995.
[18] J. J. Collins, C. J. De Luca, A. E. Pavlik, S. H. Roy, and M. S. Emley, “The effects of spaceflight on open-loop and
closed-loop postural control mechanisms: human neurovestibular studies on SLS-2,” Exp Brain Res, vol. 107,
56
Chapter 3
no. 1, pp. 145-150, 1995.
[19] J. J. Collins and C. J. De Luca, “The effects of visual input on open-loop and closed-loop postural control
mechanisms,” Exp Brain Res, vol. 103, no. 1, pp. 151-163, 1995.
[20] L. Chiari, A. Cappello, D. Lenzi, and C. U. Della, “An improved technique for the extraction of stochastic
parameters from stabilograms,” Gait Posture, vol. 12, no. 3, pp. 225-234, Dec.2000.
[21] J. J. Collins, C. J. De Luca, A. Burrows, and L. A. Lipsitz, “Age-related changes in open-loop and closed-loop
postural control mechanisms,” Exp Brain Res, vol. 104, no. 3, pp. 480-492, 1995.
[22] C. A. Laughton, M. Slavin, K. Katdare, L. Nolan, J. F. Bean, D. C. Kerrigan, E. Phillips, L. A. Lipsitz, and J. J. Collins,
“Aging, muscle activity, and balance control: physiologic changes associated with balance impairment,”
Gait Posture, vol. 18, no. 2, pp. 101-108, Oct.2003.
[23] L. Rocchi, L. Chiari, A. Cappello, A. Gross, and F. B. Horak, “Comparison between subthalamic nucleus and
globus pallidus internus stimulation for postural performance in Parkinson’s disease,” Gait Posture, vol. 19,
no. 2, pp. 172-183, Apr.2004.
[24] T. E. Prieto, J. B. Myklebust, R. G. Hoffmann, E. G. Lovett, and B. M. Myklebust, “Measures of postural steadiness:
differences between healthy young and elderly adults,” IEEE Trans Biomed Eng, vol. 43, no. 9, pp. 956-966,
Sept.1996.
[25] S. J. Olney and D. A. Winter, “Predictions of knee and ankle moments of force in walking from EMG and
kinematic data,” J Biomech, vol. 18, no. 1, pp. 9-20, 1985.
[26] L. Chiari, L. Rocchi, and A. Cappello, “Stabilometric parameters are affected by anthropometry and foot
placement,” Clin Biomech (Bristol. , Avon. ), vol. 17, no. 9-10, pp. 666-677, Nov.2002.
[27] L. Chiari, A. Bertani, and A. Cappello, “Classification of visual strategies in human postural control by
stochastic parameters,” Hum Mov Sci, vol. 19, pp. 817-842, 2000.
[28] P. Rougier, “Influence of visual feedback on successive control mechanisms in upright quiet stance in
humans assessed by fractional Brownian motion modelling,” Neurosci Lett, vol. 266, no. 3, pp. 157-160,
May1999.
[29] T. Mergner, C. Maurer, and R. J. Peterka, “A multisensory posture control model of human upright stance,”
Prog Brain Res, vol. 142, pp. 189-201, 2003.
[30] N. L. Cerminara, A. L. Edge, D. E. Marple-Horvat, and R. Apps, “The lateral cerebellum and visuomotor control,”
Prog Brain Res., vol. 148:213-26., pp. 213-226, 2005.
[31] J. J. Buchanan and F. B. Horak, “Voluntary control of postural equilibrium patterns,” Behav. Brain Res, vol.
143, no. 2, pp. 121-140, Aug.2003.
[32] A. M. Wong, M. Y. Lee, J. K. Kuo, and F. T. Tang, “The development and clinical evaluation of a standing
biofeedback trainer,” J Rehabil Res Dev, vol. 34, no. 3, pp. 322-327, July1997.
57
Direction Specificity of Audio-Biofeedback for Postural Sway
58
Chapter
Chapter 4
Direction Specificity of AudioBiofeedback for Postural Sway
59
Direction Specificity of Audio-Biofeedback for Postural Sway
60
Chapter 4
Abstract
Sway reduction induced by use of biofeedback devices has been widely documented
in postural control research. However, the extent to which subjects use a generalized versus
a direction-specific mechanism to reduce sway is unknown. In this study, we investigated
the effects of audio biofeedback related to medial-lateral trunk acceleration or to anteriorposterior trunk sway on medial-lateral and anterior-posterior center-of-pressure displacement
during stance. Results show that direction-specific, audio-biofeedback allowed subjects to
reduce their center-of-pressure displacement by increasing the frequency of their postural
corrections in the specific direction of the audio-biofeedback. The direction-specific reduction
of center-of-pressure displacement and increase of its frequency bandwidth associated with
direction-specific biofeedback found in this study suggests that subjects do not reduce
center-of-pressure displacement by a general stiffening strategy but by increasing closedloop control of posture.
61
Direction Specificity of Audio-Biofeedback for Postural Sway
Introduction
Different mechanisms have been suggested to reduce sway in subjects attempting to
control their stance under a variety of experimental conditions. Sway reduction during stance
has been reported to be due to (1) a noise reduction in sensory feedback loop associated with
an increase in availability of sensory information [1], (2) an enhanced feedforward control from
repetitive balance training [2], (3) a generalized cognitive interference from the performance
of a dual task [3], and (4) a change in postural alignment and generalized muscle stiffness
associated with a threat of a fall [4]. Understanding the mechanisms used by subjects to
reduce their sway under different conditions is fundamental in order to determine how the
central nervous system is involved in this process.
The mechanisms applied by subjects to reduce postural sway when using biofeedback
have not been investigated. A better understanding of how the central nervous system
uses artificial sensory information to reduce postural sway can be exploited to improve
biofeedback systems. In this paper, we argue that sway reduction (in terms of center-ofpressure displacement and acceleration at trunk level) associated with audio biofeedback
related to direction of postural sway is not the consequence of a simple, generalized
mechanisms but rather the consequence of an increase in active, directionally-specific neural
control of postural stability.
62
Chapter 4
Materials & Methods
Eight healthy adults participated in this study (22-44 years old, 4 females and 4 males,
age 33±7 years, weight 71±16 kg, and height 175±11 cm). The subjects were divided into 2,
4-person groups and were gender- and age-matched between the groups. Subjects were
excluded if they reported: a use of medications and/or a history of surgeries that may have
affected their balance or their hearing, sensory loss, hearing deficits, and neurological disorders.
The rights of the participants were protected according to the Declaration of Helsinki. Each
subject signed an informed consent form in accordance with the OHSU Institutional Review
Board regulations for human subjects.
The subjects were instructed to maintain balance while standing with eyes closed on a
force plate (AMTI OR6-6) with feet 2-cm apart from each other (narrow stance). A prototype
ABF system [5] was used to provide subjects with trunk acceleration information via earphones.
The ABF system provided direction-specific information: either anterior-posterior (AP) or
medial-lateral (ML) information about the subject’s trunk movements. The AP and ML ABF
were customized for each subject and for each trial by calculating the mean and the standard
deviation (SD) of the subject’s acceleration during the first 10 seconds of each trial [6].
While the subjects’ acceleration at trunk level was inside a 2-SD range from their mean
acceleration, which was calculated in the first 10 seconds of each trial, a 400-Hz, low-volume,
pure tone was provided to the subjects in both earphones. As soon as they exceeded the
2-SD range, the stereo sound was modulated in pitch and volume in order to represent the
subject’s acceleration at trunk level, and the subjects were encouraged to adjust their sway
in order to return within the 2-SD range. The AP information was encoded by modulating
the pitch and the volume of the ABF sound. Specifically, when the subjects swayed forward
(AP acceleration increased in the anterior direction), pitch and volume increased in both
earphones whereas, when the subjects swayed backward (AP acceleration increased in
the posterior direction), pitch decreased and volume increased in both earphones. The ML
information was encoded by modulating the left/right balance of the stereo ABF sound.
Also, the ABF sound became louder the more the subject leaned far from the vertical (ML
acceleration increased). Thus, when the subjects swayed leftward (ML acceleration increased
in left direction), the volume increased in the left earphone and decreased in the right one,
and when they swayed rightward (ML acceleration increased in right direction), the volume
increased in the right earphone and decreased in the left one. The equations used to create
63
Direction Specificity of Audio-Biofeedback for Postural Sway
the ABF sound are described in detail in [5].
Before the experiment, subjects were told how to use the ABF and practiced with the
ABF system until they felt confident in performing the experiment. All subjects performed
a total of 16, 1-minute trials: 4 with the AP ABF, 4 with the ML ABF, and 8 with no ABF. Trials
alternated between those with and without ABF. The first group of subjects performed all AP
ABF trials first; the second group performed all ML ABF trials first.
Center of pressure (COP) displacement in the AP and ML directions was calculated from
the forces and torques sensed by the force plate. From the ABF system, acceleration sensed
at the trunk along the AP and ML directions was also recorded. All data were acquired with
a 100-Hz sample rate, using a NI-DAQcard 6024E and ABF custom-made software [5].
Trunk acceleration root mean square was post-processed for the AP direction and the
ML direction, and for both directions combined (RMSAP, RMSML, and RMS, respectively). Root
mean square of the acceleration was intended as an indicator of the subject s’ sway area
because it is highly correlated with the COP root mean square [5] (see figure 1A), which is
traditionally used to quantify the stability of postural sway [7]. From COP data, the frequency
comprising the 95% of the COP power spectrum [7] was post-processed for the AP and ML
directions and for the two directions combined (F95%AP, F95% ML, and F95%, respectively).
These last parameters are computed as the frequency comprising the 95% of the power of
the signal spectrum [7]. As a consequence, they are an approximation of the signal bandwidth.
An increase of these parameters suggests the power is shifting toward higher frequencies.
Under a physiological standpoint, this can be explained as an increase in the amount and
intensity of postural corrections. The mean position of COP displacement was also calculated
for each trial.
A 2-way, repeated measure, mixed, factorial ANOVA was performed on the data, with
the group (first or second) being the between factor and the ABF mode (AP, ML, off ) being
the within factor. Bonferroni post-hoc tests were performed to discriminate the effects of the
different ABF modes on the parameters extracted from the COP and the acceleration data.
Paired T-test were used to verify if mean position of COP displacement changed while subjects
used the ABF. The threshold for statistical significance was fixed at p=0.05.
64
Chapter 4
Results
Direction-specific ABF reduced subjects’ sway (in terms of center-of-pressure
displacement and acceleration at trunk level) in the specific direction of the ABF by increasing
the frequency of postural corrections in the direction of the ABF. For both AP ABF and ML ABF,
sway decreased and postural corrections increased in the direction of the feedback twice as
much as in the direction without feedback. Figure 1A shows raw AP data from COP and trunk
acceleration from one representative subject in two conditions, without ABF (dark gray) and
with AP ABF (light gray). The direction of ABF main factor was statistically significant for all
the parameters (p<0.05 for RMSML and p<0.01 for all the other parameters). However, there
was no statistical significance found for any parameters between the group that began with
the AP ABF trials and the group that began with the ML ABF trials. In addition, there was
no significant interaction found between group and ABF mode. Post-hoc analysis verified
that both AP and ML ABF significantly reduced RMS and increased F95%. In addition, AP
Figure 1 – Figure 1 – Panel A: Acceleration (top) and COP raw data (bottom) in the AP direction from a
representative subject are illustrated. The light gray lines in panel A represent the subject’s sway when using ABF;
the dark gray lines represent the subject’s sway when not using ABF. The threshold used for ABF was based on
standard deviation and it is represented as a dashed light gray line in panel A (top). Panel B: percent changes from
the condition without ABF of RMSAP, RMSML, F95%AP, and F95%ML while using ABF in the AP direction (left) and in
the ML direction (right). Both AP and ML ABF reduced sway and increased the frequency of postural corrections in
the specific direction of the ABF. (* indicates p<0.05).
65
Direction Specificity of Audio-Biofeedback for Postural Sway
ABF significantly reduced RMSAP but did not significantly influence RMSML. Similarly, ML ABF
significantly reduced RMSML but did not significantly influence RMSAP. Figure 1B shows the
averaged effect of AP and ML ABF on RMSAP and RMSML. AP ABF significantly increased F95%AP
more than F95%ML for all but one subject. Furthermore, ML ABF significantly increased F95%ML
but did not significantly influence F95%AP. Figure 1B shows the averaged effect of AP and ML
ABF on F95%AP and F95%ML. Also, the mean position of COP displacement did not significantly
change (p>0.5) when subjects used the ABF.
66
Chapter 4
Discussion
This study shows that ABF providing direction-specific information about trunk
acceleration with respect to gravity reduced subjects’ sway in the specific direction provided
by the ABF by increasing the frequency of postural corrections in that direction. In fact, for
both AP and ML biofeedback, sway parameters were affected twice as much in the direction
of ABF than in the orthogonal direction. AP ABF influenced all sway parameters more than
ML ABF probably because AP sway has a larger range of motion and consequently, a larger
tolerance for parameters that can change. In addition, the fact that F95% ML significantly
increased with AP biofeedback may have been induced by a higher activity of the TIB muscles.
In fact, a co-activation of TIB muscles to move the subject forward increases ML stiffness as
a result of the not orthogonal force exerted by these muscles. This higher ML stiffness may
have been reflected in our study by an increase of F95%ML.
There are several different mechanisms by which ABF could have influenced postural
sway but we favor a mechanism involving increased sensory feedback control [6]. It is unlikely
that sway was reduced as a consequence of generic auditory stimulation because auditory
stimulation unrelated to body sway has been found to increase, not decrease, postural sway [810]. It is also unlikely that the dual task required by attending to auditory cues while balancing
was responsible for sway reduction, because cognitive tasks usually increase postural sway
[11] and direction-specific sway reduction due to a secondary cognitive task has not been
reported [3].
ABF may be able to reduce postural sway by generalized muscle co-contraction. A
generalized co-contraction, according to an inverted pendulum model, would increase sway
area and increase sway frequency in both AP and ML direction. In our previous study with
AP and ML ABF while standing on foam, subjects reduced sway area and increased sway
frequency without increasing muscular co-contraction [6].
Lengthening and activating the Tibialis Anterior muscle can result in decrease AP but
not ML sway. Carpenter and Frank (2001) showed that subjects may decrease sway area and
increase sway frequency in AP, but not ML direction, when faced with the threat of a fall from
standing while facing the edge of a high support surface. Although ABF and the threat of
falling similarly affected the standard deviation of COP displacement (-10% from fear and -5%
with ABF) and increase the mean frequency of COP (+15% from fear and +18% from ABF), the
mechanisms differ. With threat of a forward fall, the changes in AP postural sway appear to
67
Direction Specificity of Audio-Biofeedback for Postural Sway
be a consequence of a backward shift of the mean COP which increased the magnitude and
the duration of activity in the Tibialis Anterior muscle. In our study, the mean COP position
did not significantly change with ABF. In other studies, we also found this ABF device that
ABF did not alter leg muscle activity or co-contraction, so increased muscle stiffness cannot
explain sway reduction due to ABF [6].
Postural sway can be controlled with both feedback and predictive, feedforward
mechanisms [12;13]. Using stabilogram diffusion analysis, we previously showed that the
short-term, “closed loop” component was increased whereas, the long-term, “open loop”
component was decreased by ABF [6]. Although ABF appears to reduce sway primarily via
an increase in sensory feedback control in our studies, it is possible that with more practice,
the biofeedback task may become more automatic so that subjects could rely more on feed
forward control provided by the trunk acceleration signals [2;14;15].
Direction-specific ABF was found to induce direction-specific reduction in postural sway
by increasing the frequency of postural corrections. These results are consistent with an
active integration of ABF with other sensory information by the nervous system to enhance
postural control.
68
Chapter 4
Bibliography
[1]
R. A. Speers, A. D. Kuo, and F. B. Horak, “Contributions of altered sensation and feedback responses to changes
in coordination of postural control due to aging,” Gait Posture, vol. 16, no. 1, pp. 20-30, Aug.2002.
[2]
M. C. Dault and J. S. Frank, “Does practice modify the relationship between postural control and the
execution of a secondary task in young and older individuals?,” Gerontology, vol. 50, no. 3, pp. 157-164,
May2004.
[3]
D. L. Weeks, R. Forget, L. Mouchnino, D. Gravel, and D. Bourbonnais, “Interaction between attention
demanding motor and cognitive tasks and static postural stability,” Gerontology, vol. 49, no. 4, pp. 225232, July2003.
[4]
M. G. Carpenter, J. S. Frank, C. P. Silcher, and G. W. Peysar, “The influence of postural threat on the control
of upright stance,” Exp Brain Res, vol. 138, no. 2, pp. 210-218, May2001.
[5]
L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[6]
M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A. Cappello, “Influence of a portable audio-biofeedback
device on structural properties of postural sway,” J Neuroengineering Rehabil, vol. 2, p. 13, May2005.
[7]
T. E. Prieto, J. B. Myklebust, R. G. Hoffmann, E. G. Lovett, and B. M. Myklebust, “Measures of postural steadiness:
differences between healthy young and elderly adults,” IEEE Trans Biomed Eng, vol. 43, no. 9, pp. 956-966,
Sept.1996.
[8]
S. A. Raper and R. W. Soames, “The influence of stationary auditory fields on postural sway behaviour in
man,” Eur J Appl Physiol Occup Physiol, vol. 63, no. 5, pp. 363-367, 1991.
[9]
R. W. Soames and S. A. Raper, “The influence of moving auditory fields on postural sway behaviour in man,”
Eur J Appl Physiol Occup Physiol, vol. 65, no. 3, pp. 241-245, 1992.
[10] T. Tanaka, S. Kojima, H. Takeda, S. Ino, and T. Ifukube, “The influence of moving auditory stimuli on standing
balance in healthy young adults and the elderly,” Ergonomics, vol. 44, no. 15, pp. 1403-1412, 2001.
[11] A. Shumway-Cook and M. Woollacott, “Attentional demands and postural control: the effect of sensory
context,” J Gerontol A Biol Sci Med Sci, p. M10-M16, 2000.
[12] K. H. van der, E. van Asseldonk, and F. C. van der Helm, “Comparison of different methods to identify and
quantify balance control,” J Neurosci Methods, vol. 145, no. 1-2, pp. 175-203, June2005.
[13] J. J. Collins and C. J. De Luca, “Open-loop and closed-loop control of posture: a random-walk analysis of
center-of-pressure trajectories,” Exp Brain Res, vol. 95, no. 2, pp. 308-318, 1993.
[14] H. Pashler, J. C. Johnston, and E. Ruthruff, “Attention and performance,” Annu Rev Psychol, vol. 52, pp. 629651, 2001.
[15] E. H. Schumacher, T. L. Seymour, J. M. Glass, D. E. Fencsik, E. J. Lauber, D. E. Kieras, and D. E. Meyer, “Virtually
perfect time sharing in dual-task performance: uncorking the central cognitive bottleneck,” Psychol Sci,
vol. 12, no. 2, pp. 101-108, Mar.2001.
69
Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss
70
Chapter
Chapter 5
Audio-Biofeedback Improves
Balance in Patients With Bilateral
Vestibular Loss
Most of the content fo this chapter has been published in: M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback
improves balance in patients with bilateral vestibular loss,” Arch Phys Med Rehabil, vol. 86, no. 7, pp. 1401-1403, July2005.
71
Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss
72
Chapter 5
Abstract
The extent to which subjects with loss of sensory information can substitute audio
information to control body sway is unknown. We developed an audio-biofeedback (ABF)
system to investigate its effect on postural stability during stance.
Audio biofeedback consisted of soundwaves representing 2D trunk cinematic (position,
velocity and acceleration) information. When the subject sway was outside a 1° threshold,
frequency and amplitude modulation signaled anterior-posterior trunk sway and left-right
ear volume balance signaled left-right sway. Nine subjects with bilateral loss of vestibular
function and nine age-matched control subjects attempted to use this biofeedback to
minimize postural sway in stance with eyes closed and with foam under their feet.
Balance stability was evaluated according to the following parameters: the root mean
square of (1) the center of pressure (COP) displacements and of (2) the trunk accelerations;
the COP bandwidth; the time spent by the participant within ±1° threshold from their baseline
COP position; and the mean accelerations of the trunk while the participant was swaying
outside this ±1° threshold.
Participants with BVL had significantly larger postural sway than did unaffected
participants. Those with BVL, while using ABF, decreased sway area by 23%±4.9%, decreased
trunk accelerations by 46%±9.9%, and increased time spent within ±1° sway threshold
by 195%±34.6%. In conclusion, ABF improved stance stability of participants with BVL by
increasing the amount of postural corrections.
73
Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss
Introduction
The brain relies on the visual, somatosensory (proprioceptive, cutaneous), and vestibular
systems to obtain reliable sensory information to control balance in stance [1]. The more
accurate this sensory information, the better is postural stability [2]. When head linear- and
angular-acceleration information are lost because of vestibular pathology, postural stability
in stance is compromised, particularly in environments lacking adequate visual and surface
somatosensory information [3]. Approximately 20% of the general population is affected
by a vestibular disorder. Patients with vestibular disorders suffer from poor balance, spatial
disorientation, and ataxia and they lack balance confidence, especially when other sensory
references are limited [4]. Recently, new technologies have produced inexpensive, small sensors
that transduce body-motion information normally provided by the human senses [5]. Such
sensors have been used to provide vibrotactile information for improving balance in normal
healthy individuals [6] and in people with vestibular loss [7]. This article reports on the effects
of using a new prototype audio-biofeedback (ABF) system based on accelerometric sensors.
This system uses the auditory input to provide sensory information, similar to that provided
by the vestibular system, to people with bilateral vestibular loss (BVL). It was hypothesized
that ABF sound coding of torso acceleration improves postural stability of people with BVL
because this additional information, which is closely related to otolith information, may at
least in part substitute for the lack of vestibular function that is the cause of balance deficits
in people with BVL [8;9].
74
Chapter 5
Methods
Participants
Nine individuals (4 men, 5 women) with severe BVL and 9 age- and sex-matched controls
performed the experiment. Participants used as controls had no balance deficit or a history of
surgeries that could affect their balance or hearing. People with BVL who had other pathologies
or a history of surgeries that could affect their balance or hearing were also excluded from
this study. All participants with BVL had bilaterally absent caloric responses and horizontal
vestibular ocular reflex gains between .005 and .140 for rotations at .05Hz. Diagnosis for
participants with BVL included 5 with gentomycin ototoxicity, 1 with Ramsey Hunt syndrome,
1 with autoimmune disorder, and 2 with idiopathic vestibular loss. Participants with BVL were
referred to our lab by neuro-otologists. The mean age of participants with BVL was 55 years
(range, 38–73yrs); mean height, 171cm (range, 160–193cm); and mean weight, 71kg (range,
51–115kg). The mean age of the control subjects was 55 years (range, 33–71yrs); mean height,
167cm (range, 151–180cm); and mean weight, 70kg (range, 65–86kg). All participants were
protected according to the 1964 Declaration of Helsinki and signed an informed consent
form before performing the experiments.
Procedures
All participants were instructed to stand with eyes closed and without footwear on an
AMTI OR6-6 force-plate with medium-density, 4-inch Temper foam in 2 conditions: with and
without ABF. The ABF system prototype (fig 1), which we developed, was equipped with a
small (3x3x1.5cm) sensor that detected antero-posterior (AP) and medio-lateral (ML) linear
accelerations at the trunk level when the sensor was applied to the torso near the body
center of mass. A laptop computer acquired the signals from the sensor and generated a
stereo sound encoding body-sway information. ML acceleration was encoded as the balance
between the volume in the left and right channel whereas AP acceleration was encoded
via both pitch and volume. As the participant leaned forward, the pitch increased, and as
the participant leaned backward, the pitch decreased. The volume always increased as the
participant leaned away from the vertical in all directions. For example, if participants swayed
diagonally forward and to the left, they heard a sound increasing in pitch in both ears and
becoming louder in the left ear and quieter in the right ear. The ABF system changed pitch or
volume only when participants exceeded their baseline sway by ±1° [10]. A 1-minute training
75
Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss
phase was enough for all participants to understand
the ABF representation of sway. Participants were
instructed to use the biofeedback sound during
trials, to correct their postural sway. Each participant
performed three 1-minute trials with ABF and three
1-minute trials without ABF, in random order. A
force-plate recorded the center of pressure (COP)
displacement under the feet. COP is the imaginary
point on the floor at which participants exert the
net reaction force to control balance. To quantify
postural stability, 5 parameters were calculated:
(1) the root mean square (RMS; mm) and (2) the
bandwidth (Hz) of the COP displacement [11], (3)
the time spent within ±1° sway threshold (s), (4) the
RMS of torso acceleration (mm/s2), and (5) the mean
torso acceleration outside the ±1° threshold (mm/
s2). Our hypothesis was that ABF would decrease the
postural sway (RMS of COP and of torso acceleration
in both AP and ML directions) and increase the time
spent inside the ±1° sway threshold, especially for
participants with BVL.
Figure 1 - Experimental setup. The center
of pressure (COP) displacements illustrated
are from 1 participant with BVL standing on
the foam with eyes closed, with (light gray)
and without (dark gray) ABF information
available. The smaller the dimension of the COP
displacement in the graphs, the smaller is the
participant’s sway. Consequently, the graphs
show how, using ABF, people can reduce their
postural sway. Abbreviation: Acc, acceleration.
Statistical Analysis
A 2-way, repeated-measure analysis of variance (ANOVA) was performed on each
dependent variable to determine the effect of ABF, the difference in postural sway between
subjects with BVL and controls, and the interaction between ABF and pathology. The criterion
for statistical significance was p less than .05.
76
Chapter 5
Results
Table 1 summarizes the results. Subjects with BVL had significantly larger postural
sway while standing on foam than did controls, and both groups decreased postural sway
using ABF in both AP and ML directions. Without ABF, 2 subjects with BVL were unable to
complete the trials standing on the foam with eyes closed, even after many attempts. However,
1 of them was successful in all trials when using ABF. Two other participants with BVL were
unable to remain standing during the first 2 attempts to perform a trial on the foam, eyes
closed, without ABF. However, with ABF they were able to remain standing throughout all
trials. For the participants with BVL who could stand on the foam with eyes closed, their COP
displacements were 65% larger than those of the controls and their torso accelerations were
22.6% larger than those of controls without ABF. Time spent within the ±1° threshold did
not differ statistically between the 2 groups without ABF. ABF significantly reduced postural
sway, as reflected by reductions in both torso acceleration and COP displacement in both AP
and ML directions. The significant interactions indicated that participants with BVL reduced
COP displacement and acceleration with ABF significantly more than did controls. Using ABF,
participants also increased the time spent inside a ±1° sway threshold and decreased their
sway acceleration while outside this threshold.
Table 1 - Effects of Using ABF on COP and Acceleration Parameters
Means of the parameters extracted
during quiet stance on foam
Percentage changes in postural
parameters with versus without ABF
Statistics, 2-way ANOVA
Significance (p)
Control
Mean (SD)
BVL
Mean (SD)
Control
% change (SEM)
BVL
% change (SEM)
BVL Pathology
ABF
2 Factors
Interaction
RMS
14.8 (3.9) [mm]
24.3 (8.7) [mm]
-15.9 (3.4)
-23.0 (4.9)
0.013
0.000
0.000
RMS AP
11.9 (2.7) [mm]
18.8 (7.5) [mm]
-15.4 (4.4)
-22.2 (4.4)
0.015
0.000
0.019
RMS ML
8.5 (3.4) [mm]
15.2 (4.8) [mm]
-15.0 (2.9)
-23.6 (6.1)
0.010
0.000
0.000
F95
1.58 (0.18) [Hz]
2.28 (0.81) [Hz]
9.3 (3.4)
8.4 (5.6)
0.000
0.060
0.830
RMS-Acc
65.2 (26) [mm/s2]
115 (81) [mm/s2]
-32.1 (10.3)
-46.2 (5.7)
0.002
0.000
0.005
RMS-Acc AP
54.6 (22) [mm/s2]
100 (52) [mm/s2]
-38.2 (10.9)
-49.8 (5.0)
0.042
0.000
0.002
RMS-Acc ML
32.9 (16) [mm/s2]
52.6 (16) [mm/s2]
-29.8 (10.8)
-35.5 (7.2)
0.060
0.001
0.030
Mean-Acc
40.9 (14) [mm/s2]
69.9 (29) [mm/s2]
-16.0 (5.9)
-25.3 (7.5)
0.001
0.053
0.387
Time-in -Thresh.
3.18 (2.82) [s]
1.91 (1.23) [s]
653.2 (336.9)
195.3 (20.1)
0.076
0.001
0.549
77
Audio-Biofeedback Improves Balance in Patients With Bilateral Vestibular Loss
Discussion
These results indicate that ABF reduced postural sway and was more effective for
subjects with BVL than for the control participants for most parameters during quiet stance
on the foam. Thus, sound may substitute, at least partially, for the lack of vestibular sensory
information to control postural sway in stance. Because participants significantly increased
time spent within the ±1° [10] sway threshold using the ABF, our conclusion is that sway
reduction was a consequence of additional postural control triggered by the audio information
[12].
In conclusion, these results suggest that a biofeedback system, such as ABF, may help
people with BVL improve balance when attempting to stand in environments with surface
somatosensory and visual information inadequate for postural control. Also, this ABF device
may be useful for balance training rehabilitation, as it has been found in other studies of
postural biofeedback [13-15]. Future studies are needed to determine (1) whether people,
after practicing with ABF, can use this additional information more automatically, without
focused attention on feedback or postural control and (2) whether ABF is useful for stabilizing
dynamic balance in tasks such as gait.
78
Chapter 5
Bibliography
[1]
F. B. Horak and J.M. Macpherson, “Postural equilibrium and orientation,” in Handbook of Physiology. Rowell
R.B. and Shepherd J.T., Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[2]
A. D. Kuo, R. A. Speers, R. J. Peterka, and F. B. Horak, “Effect of altered sensory conditions on multivariate
descriptors of human postural sway,” Exp. Brain Res, vol. 122, no. 2, pp. 185-195, Sept.1998.
[3]
L. M. Nashner, F. O. Black, and C. Wall, III, “Adaptation to altered support and visual conditions during stance:
patients with vestibular deficits,” J Neurosci, vol. 2, no. 5, pp. 536-544, May1982.
[4]
F. B. Horak and C. L. Shupert, “Role of the vestibular system in postural control,” in Vestibular Rehabilitation.
S. J. Herdman, Ed. Philadelphia: F. A. Davis Company, 1994, pp. 22-46.
[5]
R. E. Mayagoitia, A. V. Nene, and P. H. Veltink, “Accelerometer and rate gyroscope measurement of kinematics:
an inexpensive alternative to optical motion analysis systems,” J Biomech, vol. 35, no. 4, pp. 537-542,
Apr.2002.
[6]
C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical sensors
using vibrotactile feedback of tilt,” IEEE Trans Biomed Eng, vol. 48, no. 10, pp. 1153-1161, Oct.2001.
[7]
E. Kentala, J. Vivas, and C. Wall, III, “Reduction of postural sway by use of a vibrotactile balance prosthesis
prototype in subjects with vestibular deficits,” Ann Otol Rhinol Laryngol, vol. 112, no. 5, pp. 404-409,
May2003.
[8]
T. Brandt, Vertigo: Its Multisensory Syndromes. New York: Springer-Verlag, 1991.
[9]
A. Shumway-Cook and F. B. Horak, “Rehabilitation strategies for patients with vestibular deficits,” Neurol.
Clin, vol. 8, no. 2, pp. 441-457, May1990.
[10] R. E. Mayagoitia, J. C. Lotters, P. H. Veltink, and H. Hermens, “Standing balance evaluation using a triaxial
accelerometer,” Gait. Posture., vol. 16, no. 1, pp. 55-59, Aug.2002.
[11] L. Rocchi, L. Chiari, and A. Cappello, “Feature selection of stabilometric parameters based on principal
component analysis,” Med Biol Eng Comput, vol. 42, no. 1, pp. 71-79, Jan.2004.
[12] M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A. Cappello, “Influence of a portable audio-biofeedback
device on structural properties of postural sway,” J Neuroengineering Rehabil, vol. 2, p. 13, May2005.
[13] A. M. Wong, M. Y. Lee, J. K. Kuo, and F. T. Tang, “The development and clinical evaluation of a standing
biofeedback trainer,” J Rehabil Res Dev, vol. 34, no. 3, pp. 322-327, July1997.
[14] S. Moore and M. H. Woollacott, “The use of biofeedback devices to improve postural stability,” Phys Ther
Practice, no. 2, pp. 1-19, 1993.
[15] A. Shumway-Cook, D. Anson, and S. Haller, “Postural sway biofeedback: its effect on reestablishing stance
stability in hemiplegic patients,” Arch Phys Med Rehabil, vol. 69, no. 6, pp. 395-400, June1988.
79
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
80
Chapter
Chapter 6
Auditory Biofeedback Substitutes
for Loss of Sensory Information in
Maintaining Stance
Most of the content of this chapter was published in: M. Dozza, F. Horak, and Chiari L., “Auditory Biofeedback Substitutes
for Loss of Sensory Information in Maintaining Stance,” Exp. Brain Res, 2006.
81
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
82
Chapter 6
Abstract
The importance of sensory feedback for postural control in stance is evident from the
balance improvements occurring when sensory information from the vestibular, somatosensory,
and visual systems is available. However, the extent to which also audio-biofeedback (ABF)
information can improve balance has not been determined. It is also unknown why additional
artificial sensory feedback is more effective for some subjects than others and in some
environmental contexts than others.
The aim of this study was to determine the relative effectiveness of an ABF system
to reduce postural sway in stance in healthy control subjects and in subjects with bilateral
vestibular loss, under conditions of reduced vestibular, visual, and somatosensory inputs. This
ABF system used a threshold region and non-linear scaling parameters customized for each
individual, to provide subjects with pitch and volume coding of their body sway.
ABF had the largest effect on reducing the body sway of the subjects with bilateral
vestibular loss when the environment provided limited visual and somatosensory information;
it had the smallest effect on reducing the sway of subjects with bilateral vestibular loss,
when the environment provided full somatosensory information. The extent that all subjects
substituted ABF information for their loss of sensory information was related to the extent
that each subject was visually-dependent or somatosensory-dependent for their postural
control.
Comparison of postural sway under a variety of sensory conditions suggests that patients
with profound bilateral loss of vestibular function show larger than normal information
redundancy among the remaining senses and ABF of trunk sway. The results support
the hypothesis that the nervous system uses augmented sensory information differently
depending both on the environment and on individual proclivities to rely on vestibular,
somatosensory or visual information to control sway.
83
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
Introduction
The control of postural sway depends on continuous feedback of sensory information
from the vestibular, somatosensory, and visual senses. The largest increase in postural sway in
stance occurs when somatosensory information is compromised [1]. The next largest increase
occurs when vestibular information is lost, and the smallest, when vision is eliminated by eye
closure [2-4]. These increases in postural sway suggest that the central nervous system (CNS)
relies primarily on somatosensory information, less so on vestibular information, and even
less so on visual information to control postural sway during quiet stance. In fact, a linear
sensory interaction model predicts such postural sway in adults during stance by proposing
a 70% dependence on somatosensory information from a firm surface, 20% on vestibular
information, and 10% on visual information [5]. However, several studies support the notion
that the CNS re-weighs its relative dependence on sensory information when the availability
of information from different senses changes [6-8]. For example, when healthy subjects stand
on an oscillating surface with eyes closed, they increasingly depend on vestibular information
and visual information and decrease dependence on somatosensory information from the
surface as the amplitude of the surface rotations increases [5].
It is as yet unknown the extent to which the CNS reweighs its relative dependence on
sensory information in presence of augmented sensory information. Augmentation of sensory
information, such as auditory information, could be useful for rehabilitation of balance in
patients with sensory loss, especially if the CNS proportionately integrates this information
with the natural sensory information depending on the sensory demands of the task.
One type of augmentation to reduce postural sway—auditory information in the form
of biofeedback—has received minimal investigation. When audio-biofeedback (ABF) was
investigated, it was usually in conjunction with visual biofeedback [9;10]. In studies of ABF
and visual biofeedback, the sound constituting the ABF was a simple alarm signal [11;12]
that was used to augment the visual biofeedback. However, another type of ABF, able to
represent a complex information and not limited to an alarm signal, may be especially useful
to augment postural feedback since auditory cues: (1) are easy to integrate with the remaining
senses in sensory-impaired individuals, such as those with vestibular losses [13], (2) do not
interfere with visual information, and (3) are capable of signaling spatial information [14;15].
To illustrate this last point, humans use hearing for spatial localization whenever we turn our
heads to locate the source of a sound. In addition, it has been shown that novice pilots can
84
Chapter 6
learn how to fly in a flight simulator using either visual information or auditory tracking for
turns, bank angles, and tilt [16], and it was subsequently determined that healthy subjects can
use auditory information nearly as accurately as visual information to detect body orientation
and motion in space [14].
Auditory and vestibular information are both transmitted to the brain via the VIII cranial
nerve, which projects to the temporal lobe. Auditory cues automatically (subconsciously)
influence postural alignment, and postural alignment automatically alters the ability to locate
auditory cues in the environment [17;18]. Even stationary auditory cues were found to reduce
the body sway of control and blind subjects when the cues were from stereo speakers in
close proximity to both ears [19].
Recently, it has been found that subjects with a loss of vestibular information were able
to use both ABF [20;21] and tactile biofeedback [22;23] that map their body movement in
order to reduce postural sway. However, subjects with and without vestibular loss varied widely
in their ability to reduce sway with augmented sensory ABF and vibrotactile biofeedback. The
reasons for this inter-subject variability are unknown. However, similar inter-subject variability
was also found when subjects with and without vestibular loss relied on their three natural
sources of sensory information (visual, vestibular, and somatosensory) to control postural sway
[7;24]. For example, 50% of subjects with neuromas on the VIII cranial nerve increased their
postural sway in stance with eyes closed, but 50% decreased or did not change their sway
with eyes closed [24]. After surgery to remove the neuroma, the same subjects, who were
visually dependent (i.e., relied more on visual than on somatosensory information to maintain
balance) before the surgery, no longer increased their sway with eyes closed, whereas those
subjects who were not visually dependent increased their sway with eyes closed after surgery.
Further, as people age or are exposed to weightlessness in space for a long time, many, but not
all, increased their relative dependence on visual and somatosensory information to maintain
balance [25;26]. Sensory compensation for pathological loss of sensory information has also
been found to vary among subjects with profound bilateral loss of vestibular information (BVL,
bilateral vestibular loss). Fifty percent of these subjects were able to significantly reduce body
sway during surface oscillations by opening their eyes, whereas the other fifty percent could
not [27]. Studying BVL subjects using a custom-made ABF, Hegeman [21] reported balance
improvements when they stood with eyes open on firm surface but not on foam surface or
with eyes closed. However, Hegeman [21] did not perform any analysis aimed at understanding
how and why individual subjects were able or unable to use the ABF information to improve
their stability in the different postural tasks.
In the study described here, we investigated how individual subjects’ relative dependence
on a particular sensory channel influenced their ability to reduce postural sway in stance
when they used ABF information to control body sway. The objectives of this research were
(1) to determine the extent to which ABF information helps control postural sway given
85
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
limited visual, vestibular, and surface somatosensory information and (2) to account for why
the relative effectiveness of ABF varies among individuals across sensory environments. We
used an ABF system, which we designed to mimic aspects of otolith vestibular information
by monitoring accelerations in the transverse plane [28].
86
Chapter 6
Methods
Participants
Nine subjects, four men and five women, with profound BVL and nine age- and gendermatched, healthy control subjects participated in this study. There were no significant age,
height, and weight differences (p > 0.05) between the BVL and control subjects, respectively:
age 55 years (38–73) versus 55 years (33–72); height 171 cm (160–193) versus 167 cm (151–
180); and weight 71 kg (51–115) versus 73 kg (65–86). Table 1 summarizes the BVL subjects’
pathologies, ages, duration of their vestibular loss, and their horizontal vestibulo-ocular reflex
(VOR) gain at 0.05 Hz. Normal VOR gains range from 0.7 to 1 for the control subjects. All BVL
subjects had a bilaterally absent response to warm and cold water on caloric tests and a VOR
gain of less than 0.3 across a range of oscillations between 0.01 and 0.1 Hz, indicating severe
loss of vestibular function [2]. In addition, each BVL subject fell without an apparent postural
response soon after the start of surface sway-referencing trials with eyes closed, consistent
with their BVL [1]. All of the BVL and control subjects were free of hearing, orthopedic, and
neurological diseases or disorders, except the vestibular pathology for BVL subjects. Written
informed consent was obtained from all subjects prior to their participation. The rights of the
participants were protected according to the 1964-Declaration of Helsinki.
Apparatus
For all experiments, the BVL and the control subjects wore a custom-made ABF system
[28] while standing on an AMTI OR6–6 force plate. The ABF system provided auditory
information to the subjects about their body sway while they stood on the force plate.
The ABF system is comprised
of three main parts: the sensory unit,
the sensory processing unit, and the
audio output unit [28]. The sensory
unit consists of a small (1.5 x 3 x 3
cm 3 ) sensor that is mounted on the
subject’s back at L5 with a Velcro
belt. The sensory unit uses 3031
Eurosensor accelerometers (range
±0–50, resolution 2 x 10 -4 g, noise
Table 1 – Characteristics of BVL Subjects including the vestibuleocular reflex gain (VOR)
Subject ID
Age
1
46
2
50
3
56
4
60
5
61
6
38
7
53
8
56
9
73
Diagnosis
Duration of Loss
VOR
Otoxicity
7
0.030
Idiopathic
12
0.006
Idiopathic
14
0.005
Ramsey Hunt
3
0.020
Otoxicity
10
0.047
Auto Immune Desease
7
0.140
Otoxicity
7
0.260
Otoxicity
10
0.007
Otoxicity
9
0.022
87
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
1 μv p-p, temperature error-zero -0.05 mV/°C) to sense the linear accelerations along the
anterior–posterior (AP) and medial–lateral (ML) directions near the body center of mass.
L5 was chosen because its position is minimally affected by movement artifacts, such as
respiration, heartbeat, and voluntary head or limb movement. The processing unit consists of
a laptop computer (Intel Celeron 2.4 GHz) equipped with an A/D board (DAQCard NI 6024E).
It acquires, records, and processes the AP and ML accelerations sensed by the sensory unit
and encode them into two analog sine waves that constitute the ABF stereo sound. The
closed-loop delay introduced by the processing was estimated to be 5 ms. We developed the
software for the processing unit using Matlab© 6 R12 and Matlab Data Acquisition Toolbox
[28]. The audio output unit consists of an amplifier (Fostex PH- 5) that boosts the two sine
waves provided by the computer so that the subjects are able to hear tones through the
earphones (Philips SBC HP-140), with the tones representing the degree and direction of the
body accelerations.
The force plate estimates body sway in the AP and ML directions by recording forces
and torques under the subject’s feet. In certain testing conditions, the force plate was covered
with a 10 cm-thick, medium density TemperTM foam (indentation force deflection at 25%: 116
N, tensile strength: 125 kN/m2, elongation: 109%, when temperature is 22.2°C and relative
humidity is 50%) to reduce somatosensory information about body sway from the feet.
When a subject stands on the foam, the distance between the subject’s feet and the force
plate continuously changes due to the compliance of the foam itself. As a consequence, the
estimation of the center of pressure (COP) displacement was theoretically not as accurate as
without foam. However, the error of estimation was calculated in post-process and found to
be smaller than 10%. Linear accelerations from the sensory unit, as well as forces and torques
from the force plate, were acquired with a 100-Hz sample rate.
Figure 1 shows, from a top-down perspective, four directions of sway and the relative
ABF stereo sound changes in each earphone, for each direction. The ABF left–right balance
and the volume in the earphones change according to ML body sway, and the pitch and
volume of the stereo sounds change according to AP body sway [28]. In this study, all sounds
were dynamically adjusted for each subject based on unique definitions of: (1) the region
of natural sway [29], and (2) the area of the support base that is the region of a safe sway.
Using an inverted-pendulum model [30], the region of natural sway and the region of safe
sway were uniquely calculated for each subject. Specifically, the region of natural sway was
determined by the range of AP and ML accelerations compatible with an oscillation of ±1°
around the vertical, which depended upon the subject’s height. The region of the safe sway
was determined by the range of AP and ML accelerations compatible with the subject’s COM
projection on the ground, not exceeding the subject’s base of foot support. Thus, the region
of natural sway and region of safe sway were used to customize and to optimize the ABF
tones for each subject.
88
Chapter 6
Figure 1 – ABF sound dynamics encoding postural sway. Pitch and volume change in the two earphones,
depending on the direction of sway. The arrows in the middle of the force plate (outlined) indicate the direction of
sway. The regions of natural sway (NS) and safe sway (SS) were customized for each subject.
The ABF system was designed so that the tones changed, depending on the subject’s
sway relative to the calculated region of natural sway. When a subject swayed within his
or her calculated region of natural sway in the ML and AP directions, the same constant,
low-volume (20 dB-SPL), 400-Hz tone was fed back to the subject through each earphone.
However, when a subject swayed outside his or her region of natural sway in the ML direction,
the tones in the earphones simultaneously became louder in the ear corresponding to the
direction of body sway and quieter in the other ear. When the subject swayed outside the
region of natural sway in the anterior direction, the tones changed equally in both ears and
became louder (up to 50 dB-SPL) in volume and higher in pitch (following a linear function
up to 1000 Hz). When the subject swayed outside the region of natural sway in the posterior
direction, the tones changed equally in both ears and became louder in volume and lower
in pitch (following a linear function down to 150 Hz). When the subject swayed outside the
region of natural sway in an oblique direction, for example in the anterior-left direction, the
tones became higher in pitch in both ears, louder in volume in the left ear, and quieter in
volume in the right ear. All the equations used to generate the ABF sound using sigmoidal
function are reported in detail in Chiari et al. [28].
Procedure
Subjects stood on the force plate and kept their feet 15° externally rotated and their
heels 1 cm apart (narrow stance position). They were instructed to maintain quiet stance
throughout all testing when using and not using the ABF device. Before the experimental
protocol began, subjects practiced with the ABF system for a few minutes on a firm surface
with eyes open by voluntarily swaying at different angles and directions, and listening to
the corresponding changes in tones in the earphones until they understood how the trunk
89
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
information was coded into the ABF sound. The subjects were instructed to correct their
body sway by using the tones, i.e., to maintain their sway within the region of natural sway
by achieving a constant 400-Hz tone in each earphone. Once they understood how to change
their body sway to achieve the constant 400-Hz tone, they performed three practice trials
with eyes closed and without ABF, followed by three practice trials with eyes open on foam
and without ABF. The purpose of the practice trials was for the subjects to gain confidence in
standing with eyes closed or standing on the foam-covered force plate without falling, and
to minimize the initial effects of standing on the foam. Data from these practice exercises
and trials were not considered in the analyses.
BVL subjects repeated a block of six conditions three times (18 trials total), and the
control subjects repeated the same block of six conditions five times (30 trials total). For each
of these blocks, the six conditions were presented in random order; three conditions were
with and three conditions were without ABF. Conditions one and two were: eyes closed on a
firm surface without ABF and with ABF. Conditions three and four were: eyes open on foam
surface without ABF and with ABF. Conditions five and six were: eyes closed on foam surface
without ABF and with ABF. We did not test the eyes-open on firm-surface condition since, in
this condition, the sway of both the BVL and the control subjects is expected to be inside
the region of natural sway, so there is no need for additional ABF information [1]. The BVL
subjects performed fewer trials to limit fatigue. Each trial lasted 1 min.
Data and statistical analysis
From the 2D, planar COP displacement, we quantified postural sway with two independent
parameters [31-33]: the root-mean-square distance (COP-RMS) and the frequency below which
the 95% of the power of the signal is included (F95%). From the 2D, planar acceleration
measured by the sensory unit, we computed the RMS (Acc-RMS). To determine the effect on
sway of subject groups, conditions, and ABF, we performed a three-way ANOVA, 2 groups
(BVL and control) x 3 sensory conditions (vestibular, somatosensory, and visual), repeated
(eyes closed, eyes open on foam, and eyes closed on foam) x 2 ABF conditions, repeated (ABF
on and off ) for each parameter (COP-RMS, F95%, and Acc-RMS). The threshold for statistical
significance was p = 0.05.
To evaluate the correlation between severity of vestibular loss and the effect of ABF on
sway amplitude in the eyes closed on foam condition, a robust regression correlation analysis
was performed between the VOR gain and the percentage reduction in COP-RMS, with and
without ABF for BVL subjects. To assess whether ABF was effective in helping subjects reduce
body sway in proportion to each subject’s level of dependency on visual and somatosensory
information, a robust regression correlation analysis was performed between the levels of
sensory dependency and the effect of ABF on COP-RMS when only visual (eyes open on foam
90
Chapter 6
with ABF condition) or only somatosensory information (eyes closed with ABF condition) was
available. The levels of visual dependency and somatosensory dependency were estimated
for each subject as the percentage of the body sway reduction occurring when visual or
somatosensory information was added (visual information, in the eyes open on foam condition
and somatosensory information, in the eyes closed condition) and were compared to the
reference eyes closed on foam condition (when neither visual and somatosensory information
was available).
91
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
Results
Center of pressure displacement
For BVL and control subjects, body sway increased as natural sensory information or
ABF information became absent or unreliable. Further, COP-RMS was significantly larger in
the eyes open on foam condition than in the eyes closed condition (p<0.05). COP-RMS was
also significantly larger when eyes were closed than when eyes were open while subjects
stood on foam without ABF (p<0.01). In the eyes closed, eyes open on foam, and eyes closed
on foam conditions, BVL subjects’ COP-RMS was significantly larger than the control subjects’
COP-RMS (p<0.001). Figure 2 shows the anterior–posterior versus lateral COP displacements
of one representative BVL subject (Fig.
2a) and one representative control
subject (Fig. 2b), in all six conditions.
Table 2 reports the COP-RMS values
in the eyes closed, eyes open on foam,
and eyes closed on foam conditions
for both subject groups.
In the three ABF conditions, both
groups benefited from ABF. That is,
ABF significantly decreased COPRMS
for both the BVL and control groups
(p<0.05). The percentage of changes
in COP-RMS due to ABF is shown in
Table 3. No significant interaction was
found between the groups and the
conditions tested since COP-RMS was
larger in BVL subjects than in control
subjects in every condition. In addition,
there was no significant interaction
between the groups and ABF as both
groups improved in the conditions
tested. A significant interaction was
found between the condition factor
92
Figure 2 – ABF sound dynamics encoding postural sway. Pitch
and volume change in the two earphones, depending on the
direction of sway. The arrows in the middle of the force plate
(outlined) indicate the direction of sway. The regions of natural
sway (NS) and safe sway (SS) were customized for each subject.
Chapter 6
Table 2 – Mean values and standard deviation (in parenthesis) of postural parameters for bilateral vestibular
loss (BVL) and control subjects in the three conditions tested without audio-biofeedback (ABF). Root mean square
distance (RMS) is reported for the center of pressure displacement (COP) and for the acceleration sensed at trunk
level (Acc). Also, the values of frequency, below which the 95% of the power of the COP signal is included, are
reported.
Parameter
Eyes closed
Eyes open on foam
Eyes closed on foam
BVL
Control
BVL
Control
BVL
Control
13.82 (8.9)
8.31 (2.8)
14.01 (9.7)
9.34 (1.2)
24.66 (7.58)
14.92 (3.7)
F95% (Hz)
1.85 (0.55)
1.31 (0.15)
1.87 (0.52)
1.39 (0.19)
2.51 (0.31)
1.59 (0.18)
Acc-RMS (mm/s2)
14.12 (8.07)
12.61 (2.4)
16.79 (9.50)
13.48 (1.9)
56.09 (19.13)
21.84 (5.60)
COP-RMS (mm)
and the ABF factor (p<0.001) due to ABF decreasing COP-RMS more in the eyes closed on
foam condition than in the eyes closed or eyes open on foam condition (Table 3). For the BVL
subjects in the eyes closed on foam condition, a significant interaction was found among all
three ANOVA factors (p<0.001) due to ABF decreasing COP-RMS the most in the eyes closed
on foam condition for all BVL subjects.
Figure 3 shows the average COP-RMS reduction when BVL and control subjects used ABF
on foam with eyes closed. As shown in Fig. 3, all but one of the BVL subjects able to perform
the eyes closed on foam condition
benefited from ABF in this condition.
In addition, BVL subject #2 fell a few
times in the eyes closed on foam
condition, but she never fell in this
condition while using ABF. BVL
subject #1 fell consistently in the
eyes closed on foam condition but
also never fell in this condition while
using ABF. BVL subject #8 benefited
from ABF, although minimally when
compared to the other BVL subjects.
BVL subject #5 (Fig. 3) was not able
to stand in the eyes closed on foam
condition, with or without ABF,
although he benefited from ABF in
the other conditions (eyes closed
and eyes open on foam). Also as
shown in Fig. 3, all control subjects
Figure 3 – The percentage of COP-RMS reduction using ABF is
benefited from ABF in the eyes
reported for each bilateral vestibular loss (a) and control (b)
subject in the condition eyes closed on foam. Data were ordered
closed on foam condition.
Frequency spectrum
For BVL and control subjects,
by percentage improvement using ABF. Subject numbers indicate
matching subjects between the groups. † BVL Subject 2 fell twice
without ABF but never fell during trials using ABF. ‡ BVL Subject 9
fell repeatedly with and without ABF. § BVL Subject 1 could stand
only with the help of ABF. Black, dashed lines represent the mean
reduction using ABF. Gray, shadowed areas represent the standard
error of the reduction using ABF.
93
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
Table 3 – Mean percentage difference of each postural parameter with and without audio-biofeedback (ABF) for
bilateral vestibular loss (BVL) and control subjects. Root mean square distance (RMS) is reported for the center
of pressure displacement (COP) and for the acceleration sensed at trunk level (Acc). Also, the values of frequency,
below which the 95% of the power of the COP signal is included, are reported.
Parameter
COP-RMS (mm)
Eyes closed
Eyes open on foam
Eyes closed on foam
BVL
Control
BVL
Control
BVL
Control
-3.24
-10.87
-9.98
-5.42
-23.07
-15.90
F95% (Hz)
21.90
23.01
10.54
18.89
8.38
9.28
Acc-RMS (mm/s2)
-20.82
-35.24
-27.38
-40.56
-46.18
-32.15
the amount of postural corrections (indicated by the parameter F95%) decreased as natural
sensory information became available or reliable and increased when ABF information was
available. Specifically, the frequency spectrum components of the COP were significantly
affected by the different test conditions, with the power at the higher frequencies increasing
when visual and/or somatosensory sensory information was reduced (p<0.001). F95% was
higher in the eyes closed on foam condition than in the eyes open on foam condition (p<0.05),
and higher in the eyes open on foam condition than in the eyes closed condition (p<0.05).
F95% was also higher for the BVL subject group than for the control group in all conditions
(p<0.001). Table 2 reports F95% values in the three conditions tested without ABF for the BVL
and control subjects. The use of ABF significantly increased F95% for both the BVL and control
subjects in all conditions (p<0.001). Table 3 shows the percent of increase in F95% when
controls and BVL subjects used ABF in each condition. There was a significant interaction
(p<0.05) between the condition tested and the presence of a vestibular deficit, with F95%
increasing in the BVL subject group more than in the control group, particularly in the eyes
closed on foam condition.
Sensory substitution
Subjects benefited from ABF information in relation to the lack of natural sensory
information. For most BVL subjects, the extent that they reduced their body sway with ABF
in the eyes closed on foam condition correlated with the extent of their vestibular loss (r =
0.76; p<0.05). Table 1 shows the VOR gains and percentage of improvement in sway for all of
the subjects using ABF. One subject with very low VOR gain (#9) could only stand with the
ABF in this condition so the percentage of improvement could not be calculated.
For both the BVL and control groups, the effectiveness of ABF in reducing body sway
was related to how dependent each subject was on visual or somatosensory information, but
not on the amount of sway in the baseline eyes closed on foam condition. Somatosensorydependent subjects benefited the most from ABF when somatosensory information
was missing, and vision-dependent subjects benefited the most from ABF when visual
information was missing. Figure 4 shows the relative dependence of each subject on visual or
somatosensory information versus the amount of benefit that each received from ABF under
conditions in which visual or somatosensory information was limited (i.e., the eyes closed
94
Chapter 6
and eyes open on foam conditions). A linear relationship for the BVL subjects and the control
subjects was found between the degree of benefit from ABF and their dependence on visual
and somatosensory information, shown by the greater number of circles in the top-right and
bottom-left quadrants of Fig. 4. The circles in the top-right quadrant represent the subjects
who were somatosensory-dependent and benefited the most from ABF when somatosensory
information was missing. The circles in the bottom-left quadrant represent subjects who were
vision-dependent and benefited the most from ABF when visual information was missing.
Figure 4 – Subjects in terms of their vision and somatosensory dependency. There is a correlation between the
use of ABF in the eyes closed and eyes open on foam conditions, and visual and somatosensory dependency. Each
subject’s tendency to rely, more on vision or somatosensory information is reported on the horizontal axis. Negative
values imply a dependency on vision more than on somatosensory information, whereas positive values imply a
dependency on somatosensory more than on vision information (a zero value on the horizontal axis indicates a
subject who relies on vision as much as on somatosensory information to maintain balance in stance). The vertical
axis shows the effect of ABF for each subject. Positive values imply ABF reduces sway more when somatosensory
information is made unreliable by standing on foam, negative values imply ABF reduces sway more when visual
information is missing (a zero value on the vertical axis indicates a subject who, when using ABF, reduces sway
when vision information is limited as much as when somatosensory information is inadequate). The Pearson
coefficient for the regression line is r=0.57 comprising data from both group and is statistically significant (p<0.05).
The Pearson coefficients reported in the figure for the two groups of subjects separated (r=0.62 and r=0.65 for
bilateral vestibular loss and control subjects, respectively) are not statistically significant (p>0.05), however they are
close to statistical significance p=0.06
95
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
Discussion
ABF efficacy in reducing sway is related to the availability of sensory information
Results from this study show that the amount that ABF compensates for missing sensory
information depends on the extent of sensory loss. When somatosensory information was
reduced (the eyes open on foam condition) and the more that BVL and control subjects were
somatosensory-dependent, the more they benefited from ABF and were able to reduce their
sway. When visual information was not available (the eyes closed condition) and the more
that BVL and control subjects were visually dependent, the more they also benefited from
ABF and were able to reduce their sway. When both somatosensory information and visual
information were limited (the eyes closed on foam condition), both BVL and control groups
showed the most benefit from ABF. Thus, we hypothesize that the degree to which subjects
benefit from ABF to reduce postural sway depends on their degree of visual, somatosensory
and vestibular loss [24;34]. Our results also showed a trend in which the more severe the
vestibular loss, the more subjects benefited from ABF. This trend needs further testing with
more subjects in order to show statistical significance. Our findings are consistent with other
studies that also reported that control and BVL subjects were able to reduce postural sway
with visual, tactile, and audio-biofeedback [21;35]. However, our study, for the first time, has
identified a potential relationship between benefits from ABF information and dependency
on sensory information.
Both BVL and control subjects’ postural sway increased when sensory information was
limited, confirming the commonly held hypothesis that the control of postural sway depends
on the amount of available sensory feedback that is available [5;36;37].
Our BVL subjects showed significantly larger sway than did our control subjects in all
conditions tested, in agreement with other studies [7;38;39]. However, the BVL subjects’ degree
of sway reduction via ABF when either visual information or somatosensory information was
available was not related to the extent of their vestibular loss. This finding may be due to the
subjects’ hesitance to rely on novel sensory information (available via ABF) when ordinary
sensory information normally and extensively used to compensate the loss of vestibular
information [24] was also available. However, this finding may also be explained by the ABF
information not yet being integrated with the subjects’ existing somatosensory and visual
information since they used ABF for only 15 min or less during testing. This lack of integration
96
Chapter 6
is also supported by another study in which we found that the use of ABF requires a larger
number of rapid postural corrections [40]. Lack of integration may be the consequence of
the subjects’ paying excessive attention to the ABF, thus interfering with the attention paid
to other sensory information. It has been shown how dual-task interference decreases with
practice over time when tasks become quasi-automatic [41]. Consequently, it may be possible
for ABF information to become more integrated with other sensory information as when ABF
is used after a longer period of time than just the few minutes in our study [42].
Attention to natural sensory information may have limited ABF efficacy in BVL subjects
Although BVL subjects reduced their sway more than the control subjects did in the
eyes closed on foam condition, they did not in the eyes open or in the firm surface conditions.
In contrast, Hegeman et al. [21] found that BVL subjects reduced sway in stance using ABF
only with eyes open on a firm surface, but not with eyes closed and/or when on foam. This
different effect of ABF may be related to differences in: (1) the design of the ABF systems,
(2) the use of trunk angular velocity instead of linear acceleration that was fed back to the
subjects, (3) the linear algorithm chosen to map trunk movement into sound, (4) subject
selection, and (5) how postural sway was measured and quantified. In our study, the high
degree of attention that BVL subjects normally pay to visual and somatosensory information
in the eyes closed and eyes open on foam conditions may have limited their ability to use ABF
since the initial use of ABF requires some a degree of attention to the tones in the earphones
[40]. Indeed, during the rehabilitation period of BVL subjects, they are taught to pay more
voluntary attention to visual and somatosensory information than would be the case if they
did not have the BVL, to compensate for the vestibular loss [43;44]. Consequently, focusing
more on visual information and somatosensory information available in the eyes open on
foam and eyes closed conditions, may have interfered with their ability to concentrate on
the ABF [45;46]. However, in the eyes closed on foam condition, when visual information and
somatosensory information were limited, subjects could focus their attention on the ABF.
Another explanation for subjects’ decreasing their sway with ABF is that their use of ABF and
the headphone equipment influenced them to pay more attention to their sway. However, in
studies in which subjects were instructed to deliberately focus their attention on their body
sway and to increase their control of posture, they did not reduce their sway [47]. Thus, we
believe that the large sway reduction induced by ABF in BVL subjects was not likely only due
to the subjects’ paying more attention to their sway.
Use of ABF reduced BVL subjects’ inter-subject Variability
We found a high inter-subject variability among BVL subjects for all the parameters
analyzed, which agrees with findings from many other studies [38;39;48]. Indeed, two of the
nine subjects did not benefit from ABF in the eyes closed condition. Some of this variability
97
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
may be explained in terms of how individual BVL subjects compensate for the vestibular
loss, which is by increasing reliance on either visual or somatosensory information [24;49]. If
inter-subject variability depends on the degree of visual or somatosensory dependency, we
may expect inter-subject variability to decrease when visual information and somatosensory
information are limited (the eyes closed on foam condition). Indeed, we found a consistent
decrease in inter-subject variability in this condition, when BVL subjects exhibited relatively
smaller standard deviations (Table 2), although their sway was larger than in the eyes open on
foam and eyes closed conditions [50]. Our BVL subjects showed significantly higher frequency
of postural corrections (F95%) than did our control subjects in all conditions tested. This result
suggests that BVL subjects were using a different mode of controlling their balance than
were the control subjects [51]. However, without kinematic measures, we cannot distinguish
between ankle and hip sway strategies, as it was done by Creath et al. [51]. The higher
frequency of postural corrections that the BVL subjects exhibited may also be related to the
higher sensory noise due to the vestibular loss that BVL have compared to control subjects.
ABF redundancy with sensory information was higher for BVL than for control subjects
In order to better highlight the difference in the use of ABF information between BVL
and control subjects, we performed a meta-analysis which combined the results from BVL and
control subjects in all the condition presented in this study in terms of sensory information
redundancy using Venn diagrams. Redundancy of sensory information occurs when the
same information is provided by more than one sensory channel. Sensory integration for
balance is driven by—that is, is dependent on—redundancy of natural sensory information
from somatosensory, visual, and vestibular channels [52]. Extensive redundancy of sensory
information provides persons with a better estimate of body segment position and kinematics,
which results in smaller postural sway [53;54].
To quantify sensory redundancy among the natural sensory information and ABF,
we averaged the sway reduction occurred in the conditions tested (when natural and ABF
sensory information was available) and represented these averages using Venn diagrams.
Figure 5 shows two Venn diagrams (one for the BVL subjects and one for the control subjects)
that represent the contributions when all or some of the sensory information channels
were contributing sensory information to control sway. The size of each diagram and their
percentages represent the percent of COP sway reduction occurred from a condition in which
ABF, somatosensory, and visual information are all limited (by turning off the ABF device, by
using foam, by closing the eyes, respectively; i.e., the eyes closed on foam condition without
ABF) and a condition when only one of these information is available.
The redundancy between the ABF contribution in reducing sway and the contribution
from each of the other sensory information was larger for BVL subjects (Fig. 5a) than
for control subjects (Fig. 5b). For BVL subjects, ABF reduced sway 46% (4, 11, and 31%)
98
Chapter 6
compared to 32% (12, 7, and 13%) for
control subjects (each of the three
percentages in parenthesis is the
amount of redundancy between ABF
information and visual, somatosensory,
and both visual and somatosensory,
respectively). From these analyses,
for BVL subjects, the redundancy
Figure 5 – Subjects in terms of their vision and somatosensory
dependency. There is a correlation between the use of ABF
among somatosensory, visual, and ABF
inFig. 5 a, b In the form of Venn diagrams the contributions
of somatosensory (SOM yellow/lighter-colored circle), visual
information was higher (31%) than for
(VIS blue/darker-colored circle), and (ABF orange/dark-gray
control subjects (13%). The greater diagram) information in reducing COP-RMS during quiet stance
for bilateral vestibular loss and control subjects, respectively.
redundancy in BVL than control Percentages indicate the size of the different areas and represent
the COP-RMS reduction experienced by the subjects when
subjects suggests that compensating
that information was available. Overlapping areas represent
redundancy of information across the sensory systems.
for vestibular loss depends on more
extensive sensory redundancy between visual and somatosensory information. Figure 5
shows that ABF information can also be redundant with visual and somatosensory sensory
information, suggesting that the CNS may treat ABF information similarly to natural sensory
information. Also, since redundancy between ABF information and other sensory information
is greater for BVL subjects than for control subjects, BVL subjects may benefit more from
the ABF information than may control subjects, especially in sensory-deprived situations. In
fact, with more practice, ABF information may also facilitate a more accurate integration and
calibration of sensory information, induced by the CNS continually comparing natural sensory
information to ABF information.
The use of foam to limit somatosensory information may have limited in the accuracy
of sensory redundancy estimation. In fact, when determining the role that the somatosensory
information plays in reducing sway (Fig. 5), we did not include all somatosensory information
that the CNS received from the entire body but only the somatosensory information from
the subject’s feet which was restricted by using the foam. Even with these qualifications,
Fig. 5 provides new insight into the mechanisms of sensory redundancy and sensory reweighing during human stance. In conclusion, we found that the BVL and the control subjects
used ABF information about their trunk acceleration to control sway, in proportion to the
extent that their other sensory information was reduced. In addition, all subjects used ABF
differently, depending on their individual proclivities to rely on vestibular, somatosensory, or
visual information in order to control sway. Redundancy between sensory information from
different sensory channels and ABF information was larger in BVL subjects than in control
subjects, suggesting that ABF information may help subjects compensate for vestibular loss
by facilitating the CNS’s integration of sensory information.
99
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
Bibliography
[1]
L. M. Nashner, F. O. Black, and C. Wall, III, “Adaptation to altered support and visual conditions during stance:
patients with vestibular deficits,” J Neurosci, vol. 2, no. 5, pp. 536-544, May1982.
[2]
R. J. Peterka and F. O. Black, “Age-related changes in human posture control: sensory organization tests,” J
Vestib Res, vol. 1, no. 1, pp. 73-85, 1990.
[3]
J. M. Macpherson and J. T. Inglis, “Stance and balance following bilateral labyrinthectomy,” Prog Brain Res,
vol. 97, pp. 219-228, 1993.
[4]
F. B. Horak and Macpherson J.M., “Postural equilibrium and orientation,” in Handbook of Physiology. Rowell
R.B. and Shepherd J.T., Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[5]
R. J. Peterka, “Sensorimotor integration in human postural control,” J Neurophysiol, vol. 88, no. 3, pp. 10971118, Sept.2002.
[6]
J. Jeka, K. S. Oie, and T. Kiemel, “Multisensory information for human postural control: integrating touch
and vision,” Exp Brain Res, vol. 134, no. 1, pp. 107-125, Sept.2000.
[7]
F. O. Black and L. M. Nashner, “Vestibulo-spinal control differs in patients with reduced versus distorted
vestibular function,” Acta Otolaryngol Suppl, vol. 406, pp. 110-114, 1984.
[8]
F. B. Horak and F. Hlavacka, “Somatosensory loss increases vestibulospinal sensitivity,” J Neurophysiol, vol.
86, no. 2, pp. 575-585, Aug.2001.
[9]
D. S. Nichols, “Balance retraining after stroke using force platform biofeedback,” Phys Ther, vol. 77, no. 5, pp.
553-558, May1997.
[10] R. Barclay-Goddard, T. Stevenson, W. Poluha, M. Moffatt, and S. Taback, “Force platform feedback for standing
balance training after stroke,” Cochrane Database Syst Rev, no. 4, p. CD004129, 2004.
[11] A. M. Wong, M. Y. Lee, J. K. Kuo, and F. T. Tang, “The development and clinical evaluation of a standing
biofeedback trainer,” J Rehabil Res Dev, vol. 34, no. 3, pp. 322-327, July1997.
[12] M. Batavia, J. G. Gianutsos, A. Vaccaro, and J. T. Gold, “A do-it-yourself membrane-activated auditory feedback
device for weight bearing and gait training: a case report,” Arch Phys Med Rehabil, vol. 82, no. 4, pp. 541-545,
Apr.2001.
[13] C. D. Wickens and J. G. Hollands, Engineering Phychology and Human Performance. New Jersey: 2000.
[14] Vinge E., “Human operator for aural compensatory tracking,”, Proceedings of the 7th Annual Conference on
Maual Control (NASA SP-281), U.S. Government Printing Office, 1971.
[15] W. T. Nelson, L. J. Hettinger, J. A. Cunningham, B. J. Brickman, M. W. Haas, and R. L. McKinley, “Effects of
localized auditory information on visual target detection performance using a helmet-mounted display,”
Hum Factors, vol. 40, no. 3, pp. 452-460, Sept.1998.
[16] T. W. Forbes, “Auditory signals for instrument flying,” J Aeronautical Society, pp. 255-258, 1946.
[17] J. R. Lackner, “The role of posture in sound localization,” Q J Exp Psychol, vol. 26, no. 2, pp. 235-251,
May1974.
[18] J. R. Lackner and P. A. DiZio, “Aspects of body self-calibration,” Trends Cogn Sci, vol. 4, no. 7, pp. 279-288,
July2000.
[19] R. D. Easton, A. J. Greene, P. DiZio, and J. R. Lackner, “Auditory cues for orientation and postural control in
sighted and congenitally blind people,” Exp Brain Res, vol. 118, no. 4, pp. 541-550, Feb.1998.
100
Chapter 6
[20] M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback improves balance in patients with bilateral vestibular
loss,” Arch Phys Med Rehabil, vol. 86, no. 7, pp. 1401-1403, July2005.
[21] J. Hegeman, F. Honegger, M. Kupper, and J. H. Allum, “The balance control of bilateral peripheral vestibular
loss subjects and its improvement with auditory prosthetic feedback,” J Vestib Res, vol. 15, no. 2, pp. 109117, 2005.
[22] M. Tyler, Y. Danilov, and Bach-y-Rita P, “Closing an open-loop control system: vestibular substitution through
the tongue,” J Integr Neurosci, vol. 2, no. 2, pp. 159-164, Dec.2003.
[23] E. Kentala, J. Vivas, and C. Wall, III, “Reduction of postural sway by use of a vibrotactile balance prosthesis
prototype in subjects with vestibular deficits,” Ann Otol Rhinol Laryngol, vol. 112, no. 5, pp. 404-409,
May2003.
[24] M. Lacour, J. Barthelemy, L. Borel, J. Magnan, C. Xerri, A. Chays, and M. Ouaknine, “Sensory strategies in
human postural control before and after unilateral vestibular neurotomy,” Exp Brain Res, vol. 115, no. 2, pp.
300-310, June1997.
[25] W. H. Paloski, M. F. Reschke, F. O. Black, D. D. Doxey, and D. L. Harm, “Recovery of postural equilibrium control
following spaceflight,” Ann. N. Y. Acad. Sci., vol. 656, pp. 747-754, May1992.
[26] M. H. Woollacott, “Age-related changes in posture and movement,” J Gerontol, vol. 48 Spec No, pp. 56-60,
Sept.1993.
[27] J. J. Buchanan and F. B. Horak, “Vestibular loss disrupts control of head and trunk on a sinusoidally moving
platform,” J Vestib Res, vol. 11, no. 6, pp. 371-389, 2001.
[28] L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[29] R. E. Mayagoitia, J. C. Lotters, P. H. Veltink, and H. Hermens, “Standing balance evaluation using a triaxial
accelerometer,” Gait Posture, vol. 16, no. 1, pp. 55-59, Aug.2002.
[30] W. H. Gage, D. A. Winter, J. S. Frank, and A. L. Adkin, “Kinematic and kinetic validity of the inverted pendulum
model in quiet standing,” Gait Posture, vol. 19, no. 2, pp. 124-132, Apr.2004.
[31] T. E. Prieto, J. B. Myklebust, R. G. Hoffmann, E. G. Lovett, and B. M. Myklebust, “Measures of postural steadiness:
differences between healthy young and elderly adults,” IEEE Trans Biomed Eng, vol. 43, no. 9, pp. 956-966,
Sept.1996.
[32] L. Rocchi, L. Chiari, and A. Cappello, “Feature selection of stabilometric parameters based on principal
component analysis,” Med Biol Eng Comput, vol. 42, no. 1, pp. 71-79, Jan.2004.
[33] C. Maurer, T. Mergner, and R. J. Peterka, “Abnormal resonance behavior of the postural control loop in
Parkinson’s disease,” Exp Brain Res, vol. 157, no. 3, pp. 369-376, Aug.2004.
[34] J. Kluzik, F. B. Horak, and R. J. Peterka, “Differences in preferred reference frames for postural orientation
shown by after-effects of stance on an inclined surface,” Exp Brain Res, vol. 162, no. 4, pp. 474-489,
May2005.
[35] C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical sensors
using vibrotactile feedback of tilt,” IEEE Trans Biomed Eng, vol. 48, no. 10, pp. 1153-1161, Oct.2001.
[36] F. B. Horak, J. Buchanan, R. Creath, and J. Jeka, “Vestibulospinal control of posture,” Adv Exp Med Biol, vol.
508, pp. 139-145, 2002.
[37] R. Dickstein, C. L. Shupert, and F. B. Horak, “Fingertip touch improves postural stability in patients with
peripheral neuropathy,” Gait Posture, vol. 14, no. 3, pp. 238-247, Dec.2001.
[38] A. Hufschmidt, J. Dichgans, K. H. Mauritz, and M. Hufschmidt, “Some methods and parameters of body sway
quantification and their neurological applications,” Arch Psychiatr Nervenkr, vol. 228, no. 2, pp. 135-150,
1980.
[39] P. M. Gagey and M. Toupet, “Orthostatic postural control in vestibular neuritis: a stabilometric analysis,”
Ann Otol Rhinol Laryngol, vol. 100, no. 12, pp. 971-975, Dec.1991.
[40] M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A. Cappello, “Influence of a portable audio-biofeedback
device on structural properties of postural sway,” J Neuroengineering Rehabil, vol. 2, p. 13, May2005.
101
Auditory Biofeedback Substitutes for Loss of Sensory Information in Maintaining Stance
[41] E. H. Schumacher, T. L. Seymour, J. M. Glass, D. E. Fencsik, E. J. Lauber, D. E. Kieras, and D. E. Meyer, “Virtually
perfect time sharing in dual-task performance: uncorking the central cognitive bottleneck,” Psychol Sci,
vol. 12, no. 2, pp. 101-108, Mar.2001.
[42] M. C. Dault and J. S. Frank, “Does practice modify the relationship between postural control and the
execution of a secondary task in young and older individuals?,” Gerontology, vol. 50, no. 3, pp. 157-164,
May2004.
[43] A. Shumway-Cook and F. B. Horak, “Rehabilitation strategies for patients with vestibular deficits,” Neurol
Clin, vol. 8, no. 2, pp. 441-457, May1990.
[44] A. Shumway-Cook, F. B. Horak, L. Yardley, and A. M. Bronstein, “Rehabilitation of balance disorders in the
patient with vestibular pathology,” in Clinical disorders of balance, posture and gait. A. M. Bronstein, T. Brandt,
and M. H. Woollacott, pp. 213-220, London: Arnold, 1996.
[45] A. Shumway-Cook, M. Woollacott, K. A. Kerns, and M. Baldwin, “The effects of two types of cognitive tasks
on postural stability in older adults with and without a history of falls,” J Gerontol A Biol Sci Med Sci, vol.
52, no. 4, p. M232-M240, July1997.
[46] M. S. Redfern, M. E. Talkowski, J. R. Jennings, and J. M. Furman, “Cognitive influences in postural control of
patients with unilateral vestibular loss,” Gait Posture, vol. 19, no. 2, pp. 105-114, Apr.2004.
[47] N. Vuillerme and G. Nafati, “How attentional focus on body sway affects postural control during quiet
standing,” Psychol Res, pp. 1-9, Oct.2005.
[48] F. O. Black, C. L. Shupert, F. B. Horak, and L. M. Nashner, “Abnormal postural control associated with peripheral
vestibular disorders,” Prog Brain Res, vol. 76, pp. 263-275, 1988.
[49] G. L. Zacharias and L. R. Young, “Influence of combined visual and vestibular cues on human perception
and control of horizontal rotation,” Exp Brain Res, vol. 41, no. 2, pp. 159-171, 1981.
[50] F. O. Black, W. H. Paloski, M. F. Reschke, M. Igarashi, F. Guedry, and D. J. Anderson, “Disruption of postural
readaptation by inertial stimuli following space flight,” J Vestib Res, vol. 9, no. 5, pp. 369-378, 1999.
[51] R. Creath, T. Kiemel, F. Horak, R. Peterka, and J. Jeka, “A unified view of quiet and perturbed stance:
simultaneous co-existing excitable modes,” Neurosci Lett, vol. 377, no. 2, pp. 75-80, Mar.2005.
[52] R. Creath, T. Kiemel, F. Horak, and J. J. Jeka, “Limited control strategies with the loss of vestibular function,”
Exp Brain Res, vol. 145, no. 3, pp. 323-333, Aug.2002.
[53] A. D. Kuo, R. A. Speers, R. J. Peterka, and F. B. Horak, “Effect of altered sensory conditions on multivariate
descriptors of human postural sway,” Exp Brain Res, vol. 122, no. 2, pp. 185-195, Sept.1998.
[54] H. van der Kooij, R. Jacobs, B. Koopman, and F.C. van der Helm, “An adaptive model of sensory integration
in a dynamic environment applied to human stance control,” Biol Cybern, pp. 103-115, Feb.2001.
102
Chapter 6
103
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
104
Chapter
Chapter 7
Effects of Linear versus Sigmoid
Coding of Visual or Audio
Biofeedback for the Control of
Upright Stance
Most of the content of this chapter was published in: M. Dozza, L. Chiari, F. Hlavacka, A. Cappello, and F. Horak, “Effects
of Linear versus Sigmoid Coding of Visual or Audio Biofeedback for the Control of Upright Stance,” IEEE Trans Neural Syst
Rehabil Eng, 2006.
105
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
106
Chapter 7
Abstract
Although both visual and audio biofeedback (BF) systems for postural control can reduce
sway during stance, a direct comparison between the two systems has never been done.
Further, comparing different coding designs of audio and visual BF may help in elucidating
how BF information is integrated in the control of posture, and may improve knowledge for
the design of innovative BF systems for postural control.
The purpose of this paper is to compare the effects of linear versus sigmoid coding
of trunk acceleration for audio and visual BF on postural sway in a group of eight, healthy
subjects while standing on a foam surface.
Results showed that sigmoid-coded audio BF reduced sway acceleration more than did
a linear-coded audio BF, whereas a linear-coded visual BF reduced sway acceleration more
than a sigmoid-coded visual BF. In addition, audio BF had larger effects on reducing center
of pressure (COP) displacement whereas visual BF had larger effects on reducing trunk sway.
These results suggest that audio and visual BF for postural control benefit from different
types of sensory coding and each type of BF may encourage a different type of postural
sway strategy.
107
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
Introduction
Biofeedback (BF) systems for postural control are aimed at providing sensory
information to supplement the natural sensory information to improve human balance [1].
Experimentation with visual BF for postural control has been in progress since the 1970s [2]
and, traditionally, involved the visualization of subjects’ center of pressure (COP) displacement
on a monitor placed in front of the subjects. Using visual BF, subjects see the movement of
their COP displacement on the computer monitor and use this information to decrease their
postural sway [3]. A few studies also reported how repetitive use of visual BF may be a valid
rehabilitation or training tool for subjects with neuropathy [4], stroke subjects [5] and healthy
elderly subjects [6;7]. However, it is still uncertain whether training with biofeedback has a
carry-over effect without biofeedback [8].
Audio BF has received much less attention than visual BF. This lack of attention to audio
BF is probably due to its relative design complexity. Whereas, visual BF could be actualized with
a standard oscilloscope connected to a force plate, audio BF requires customized computer
algorithms for their coding. In the last few years, interest in audio BF for postural control has
been renewed [9;10], partially due to advances in technology for real-time processing and
movement sensing and to new trends in wireless portable devices that can be worn during
daily activities. These new BF devices are not meant to be used only in a laboratory setting,
and offer more advantages in terms of costs and portability than visual BF devices [9-12].
It is difficult to evaluate the relative merits of the different types of BF because each
one of these, new BF system has a unique, complex design. Specifically, each one uses a
different movement sensor which assesses a different aspect of the subjects’ sway. Further, this
movement information is then fed back to the subjects by using a different coding algorithm
and through a different sensory modality; Figure 1. These substantial differences in the design
make it nearly impossible to determine which different variables in the design is responsible
for different results obtained with the different BF systems; even if these results were obtained
from similar protocols.
The strategies that subjects use to alter postural sway with different types of BF are
also unknown. Postural sway can be reduced via a number of different strategies, including
1) a general stiffening via muscle co-contraction, 2) moving the body about the ankle joints
with little motion at the knees or hips (ankle strategy; [13]), or 3) moving the body about
many joints (multisegmental strategy such as the hip strategy; [13]). Nashner and colleagues
108
Chapter 7
hypothesized that vestibular or visual inputs favored a top-down hip strategy, whereas
somatosensory inputs favored an inverted pendulum-like ankle strategy [14] so visual and
auditory BF may favor different ways of reducing postural sway. A general stiffening from
co-contraction of muscles around joints due to fear of falling has been shown to reduce COP
displacement, but it is thought to be an undesirable way of reducing postural sway because
it doesn’t improve the ability to respond quickly to external perturbations [15],
Thus, the limited knowledge to date on the effect of BF on postural control does not
allow us to determine: (1) which is the optimal algorithm to code body motion into a sensory
signal for reducing postural sway, (2) whether different postural control strategies are favored
by different designs of visual or audio BF, and (3) whether and when visual or audio BF is the
more effective in controlling sway. This study starts to address these questions for the first
time, by comparing the effects on COP displacement, trunk acceleration, and muscular activity
of two designs of visual and audio BF. The results shown in this paper provide evidence that
(1) different types of coding may be optimal for visual and audio BF, and (2) visual and audio
BF may favor different postural strategies for the control of upright stance.
Sensor
(Accelerometer)
Variable
Sensed
Postural Response
Coding
(Linear/Sigmoid)
Subjects
(Healthy, Young)
Information
Coded
Modality
(Audio/Visual)
Biofeedback Information
Figure 1 – Box diagram representing the loop design of a biofeedback system for postural control and its
application. The features of the biofeedback system used in this study are reported in parenthesis for each box
109
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
Methods
Participants
Eight, healthy, young adults (6 men and 2 women) participated in this study after
providing informed consent. Average and standard deviation of age, height, and weight of
the participants were, respectively, 23±3 yrs, 173±7 cm, and 62.5±12.5 kg. All participants
indicated that they had no known neurological, orthopedic, hearing, or balance disorders.
None needed prescription glasses. The experimental protocol was approved by the OHSU
Ethics Committee and followed the recommendations of the Declaration of Helsinki for
Human Experimentation.
Apparatus
Subjects were asked to stand on an AMTI OR6-6 force plate that was covered with a
10cm-thick TemperTM foam (Indentation Force Deflection at 25%: 116N ,Tensile Strength:
125 kN/m2, Elongation: 109%, when temperature is 72F and relative humidity is 50%) while
wearing the BF movement sensor. The foam was used to alter the somatosensory information
from the bottom of the feet and its usefulness for maintaining balance. An electromyographic
(EMG) custom-made device recorded leg muscles activity. The EMG signals from the electrodes
were amplified 20000 times, band-pass filtered (71-2650 Hz), full-wave rectified, and integrated
with a 6th order Butterworth low-pass filter with a cut-off of 50Hz. EMG signals were recorded
from the Tibialis anterior (TIB), and medial Gastrocnemius (GAS) of the dominant leg.
A custom-made BF system was used to provide subjects with two different designs
of either visual or audio BF of the acceleration sensed at trunk level (L5). This acceleration
was sensed using a 2D accelerometer (Analog Device ADXL-203), low-pass filtered (50Hz)
to cut off high-frequency noise, and amplified 4.5 times. The accelerometer was mounted
on the subject’s back using a Velcro belt. Visual BF was generated in real-time based on this
processed acceleration signal. A red, 1.5-cm-wide, 5-point star, representing the instantaneous
acceleration values along AP and ML axes, was plotted on a 15-inch, LCD monitor (resolution
1024x768 pixels; Figure 2A) that was located 50-cm away from the subjects’ eyes and adjusted
to the subject’s height. The red star subtended about 1.5 degrees of visual arc.
During all trials with visual BF, subjects were instructed to keep the red star inside a
green ellipse. Anterior-posterior (AP) acceleration was represented by vertical movements of
110
Chapter 7
B.
A.
10*SD(AccML)
R
L
R
L
R
10*SD(AccML)
R
L
L
(
2
D
)
ML
6.6*SD(AccAP)
AP)
Threshold
L
R
L
BF AP
Representation
L
R
2*SD(AccML)
R
C.
R
2*SD(AccAP)
2
nstantan
neous
o
Acceleration
10*SD(AccAP)
Acceleration Trace
L
L
D.
R
R
BF ML
Representation
2*SD(AccML)
2*SD(AccAP)
Linear Coding
L
Sigmoid Coding
Anterior-Posterior Acceleration
Linear Coding
Sigmoid Coding
Medial-Lateral Acceleration
Figure 2 – Panel A: representation of visual BF, a red star (dark gray in this figure) moves on the screen
instantaneously representing the trunk acceleration. A blue (gray in this figure) trace represents the trajectory of
the acceleration. A green ellipse (black in this figure) represents the target for the subject to pursue during the
experiment. The visual BF was scaled on the SD of the acceleration in the first 10 seconds of each trial. Panel B:
schematic representation of the dynamics of the ABF sound depending on the subject’s direction of sway. The
movements of the subject in AP and ML directions induce changes in frequency and volume for the left (L) and right
(R) channels of the stereo sound. Although during the experiment, the changes in the stereo sound characteristics
were continuous, this panel shows a qualitative representation for each direction. When the subject was inside the
threshold, the L and R channel had constant frequency and the lowest volume. An anterior movement induced a
frequency and volume increase in both channels (top side), whereas a posterior movement induced a frequency
decrease and volume increase in both channels (bottom side). Also, a movement to the left induced a higher
volume in the L earphone channel (left side) whereas a movement to the right induced a higher volume in the
R earphone channel (right side). Panels C and D: linear and sigmoid codings of BF are represented along AP (C)
and ML (D) directions. For the sigmoid coding, a threshold was also implemented so that the subject could get a
feedback about his/her movement only when exceeding this threshold.
the star while medial-lateral (ML) acceleration was represented by horizontal movements of
the star. During the experiment, a blue trace showed the star trajectory over time. Standard
deviation (SD) of the AP and ML acceleration of each subject was obtained in each trial, from
the first 10 seconds of recording and used to scale the visual BF. A green ellipse was then
displayed on the screen, with its axes aligned with the monitor axes. The vertical axis of the
111
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
ellipse was equal to twice the SD of AP acceleration and the horizontal axis, to twice the SD of
ML acceleration. The ellipse subtended about 3 degrees of visual arc and was plotted on the
screen so that 1) the abscissa of its center was on the vertical axis of the monitor; and 2) the
ordinate of its center was plotted such that its distance from the upper edge of the monitor
was 1.5 times the distance from the lower edge. These plotting rules were implemented to
take into account that subjects have larger sway dynamics in the anterior direction than in
the posterior direction [16]. The distance between the center of the ellipse and the left and
right edges of the screen were 10 times the SD of the ML acceleration in the first 10 seconds
(Figure 2A), The distance between the center of the ellipse and the upper and lower edges of
the screen were 10 times and 6.6 times the SD of the AP acceleration in the first 10 seconds,
respectively (Figure 2A).
The audio BF was based on the same AP and ML trunk acceleration coding algorithms
as the visual BF. A full description of the audio BF software and hardware can be found
in [9]. Briefly, a PC laptop was used to generate a stereo sound coding the subjects’ trunk
accelerations sensed by a bi-axial accelerometer. In this study, the accelerometer was upgraded
from the one described in [9]. This new accelerometer was preferred because of its small size,
light weight, and portability. During the trials with audio BF, the subjects stood on the force
plate while wearing a pair of earphones. The stereo sound provided by the audio BF system
consisted of two sine waves, one for the left earphone and one for the right earphone. Pitch,
volume, and left/right balance of the stereo sound were modulated to represent the AP and
ML acceleration information (Figure 2B).
Specifically, the stereo sound got (1) louder in volume and higher in pitch when the
subjects swayed forward (e.g. acceleration increased in anterior direction; volume increased
from 20 to 50-dB-SPL, frequency increased from 400Hz to 1000Hz), (2) louder in volume and
lower in pitch when they swayed backward (e.g. acceleration increased in posterior direction;
volume increased from 20 to 50-dB-SPL, frequency decreased from 400Hz to 150Hz), (3) louder
in the right ear channel (volume increased from 20 to 50-dB-SPL) and lower in the left one
(volume decreased from 20 to 0-dB-SPL) when they moved to the right (acceleration increased
in right direction), and (4) louder in the left ear channel (volume increased from 20 to 50-dBSPL) and lower in the right one (volume decreased from 20 to 0-dB-SPL) when they moved
to the left (acceleration increased in left direction). The first 10 seconds of each trial were
used to scale thresholds and limits for the dynamics of the audio BF logically and numerically
equal to the one described above, in terms of green ellipse and of screen dimensions, for the
visual BF.
Two designs of both the visual and the audio BF were presented, these two designs
were obtained by using two different coding functions, logically and numerically similar for
both the BF modalities. The simplest one was a linear function (Figure 2C-D) which mapped
the acceleration into a movement of the red star on the screen (visual BF) or a pitch and/or
112
Chapter 7
volume sound modulation in the earphones (audio BF) using a fixed, constant gain. With this
coding function there was a continuous and proportional effect of movement on the visual/
audio BF. The second coding function used a variable gain, following a sigmoid law, with a
further nonlinearity due to the presence of the threshold described above, so that subjects did
not receive any feedback information while their acceleration was below the threshold (Figure
2C-D). The sigmoid coding of visual BF used equations equivalent to the one described in [9]
for the audio BF and shown in Figure 2C-D (see [9]). The sigmoid coding function introduced 2
major characteristics: 1) the feedback was given only when movement exceeded a threshold
(i.e. when it was most needed) 2) as soon as the threshold was exceeded the sigmoid function
guaranteed a very sensitive BF modulation followed by saturation.
All software for BF and signal acquisition was implemented using Matlab and its Data
Acquisition Toolbox. An analog/digital converter (NI-DAQCard 6024E) was used to record the
accelerations from the BF system sensor, the muscle activity signals from the EMG device,
and the forces and moments from the force plate. All data were sampled with a 100-Hz
frequency.
Procedure
All participants performed 30, 55-s long trials standing barefoot on foam. Subjects were
instructed to keep their feet as close as possible but without their feet or any part of their
legs touching. A few marks on the foam helped the subjects keep their foot position across
the trials. After each trial, subjects stepped off the foam surface and waited for the foam to
return to its original shape before standing on it again for the next trial. Trials were started
5-10 seconds after the subjects stood on the foam. The 30 trials consisted of five repetitions
of six conditions. These six conditions consisted of two BF modalities (audio and visual) each
one performed in 3 different modes (linear, sigmoid, off ). The off modes conditions were used
as reference conditions and consisted of trials without sound and eyes closed for the audio
BF modality and of trials with the red star moving randomly for the visual BF modality. These
two reference conditions were chosen in order to minimize the potentially misleading effects
of attention [17] and the effect of dynamic acoustic cues on sway [18]. During the reference
condition for visual BF trials, the subjects were asked to pay attention to the movement of
the star without correcting their sway based on the random visual BF. This condition was
preferred to a blank screen, because it kept subjects paying attention to the visual task. Since
it has been confirmed that paying attention to a second task may induce sway reduction in
healthy young subjects [17], this random feedback reference condition assured that visual
BF trials were not biased by the attention devoted to a visual task. The reference condition
for audio BF also was designed to minimize external phenomena which could have reduced
sway. Since Raper & Soames (1991) [18] suggested that a random BF of sound can enlarge
postural sway, a silent reference condition was chosen. Off BF conditions were announced to
113
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
the subjects so they were aware no BF information would be provided during the trial.
The order of the conditions was randomized in five repetition-blocks. Thus, the subjects
performed a full set of six conditions randomized before repeating any of them.
Subjects had their eyes closed during all audio BF modes and open during all visual
BF modes. During the trials with BF, subjects were asked to correct their sway according to
the feedback, by keeping the red star inside the green ellipse for visual BF and keeping the
volume as low and as balanced as possible for audio BF.
Data Analysis
To compare efficacy of BF to reduce postural sway, for each trial, the root mean square
(RMS) was calculated for the 2D trunk acceleration and the 2D COP displacement. The RMS
reflects extent of sway displacement [19]. These parameters were calculated according to
Prieto et al., (1996) [20] and were chosen because, according to Rocchi et al., (2004) [21] and
Maurer et al., (2005) [22] they complement each other in characterizing sway displacement.
Trunk acceleration reflects body COM acceleration because so much of the COM is in the
trunk [13], and it is highly correlated to the COP displacement [9] when subjects use and
ankle strategy to maintain balance. The COP displacement reflects body tilt as well as forces
the subject exerts into the ground to move the body COM [23]. To further determine the
effect of different types of BF on the strategies subjects use to control postural sway, the
mean activity of TIB as measured by EMG signals, and the level of co-contraction between TIB
and GAS were calculated as an indication of a stiffening strategy. According with Olney and
Winter 1982 [24], co-contraction was quantified as the correlation coefficient between the
low-pass filtered EMG signals. The AP shear force vector was measured as a reflection of the
extent of hip strategy used to correct postural sway [13;25]. Also, the correlation between
trunk acceleration and COP displacement was calculated along AP axis to determine whether
subjects were moving with an inverted pendulum (ankle) strategy (high correlations), or as
more complex, multi-segmental (hip or other) kinematics strategy, (low correlations).
Paired T-tests were used to compare the effects of the sigmoid and linear designs for
coding the visual or audio BF on the parameters above. All comparisons were made on
percent change due to BF from baseline conditions because the baseline condition without
visual and without audio BF differed (eyes open with peripheral view of the room for vision
and eyes closed for audio with significantly more sway p<0.01).
114
Chapter 7
Results
A. % Changes due to Audio Biofeedback
0
LINEAR
SIGMOID
SIGMOID
LINEAR
Effectiveness of Biofeedback
The BF induced a significant
reduction (p<0.05) in the RMS of trunk
acceleration in all but the linear coding
of audio BF condition. Linear coding of
visual BF reduced trunk acceleration
more than sigmoid BF (p<0.05). In
contrast, sigmoid coding of audio BF
reduced trunk acceleration more than
linear BF (p<0.05). Figure 3 shows
the percent changes induced by the
different modalities and coding of BF
on the RMS of the trunk acceleration.
Figure 4A shows the raw, AP trunk
acceleration data from a representative
subject while using linear and sigmoid
audio and visual BF.
*
*
-20
*
Acceleration RMS
*
COP RMS
B. % Changes due to Visual Biofeedback
20
0
LINEAR
SIGMOID
LINEAR
SIGMOID
Acceleration RMS
*
-20
COP RMS
*
-40
Figure 2 – Average percentage changes of acceleration RMS
and COP RMS while using audio BF (panel A) and visual BF
(panel B) referenced to the relative, off conditions (eyes closed
for audio BF and eyes open for visual BF). The effect of BF on
acceleration RMS is represented in white, whereas the effect
of BF on COP RAM is represented in gray. The asterisks, which
are close to brackets, indicate statistical significant difference
(p<0.05) between histograms. The asterisks, which are close
to the histograms, indicate statistical significance (p<0.05)
from the BF condition represented by the histogram and the
respective off condition.
The effect of BF on COP RMS also
depended on the BF modality and its
coding. Only the sigmoid coding of
audio BF significantly reduced COP
RMS (p<0.05). Figure 3B shows the percent change of COP RMS induced by the different
modalities and coding of BF. Figure 4B shows the raw, AP COP data from a representative
subject while using sigmoid and linear audio and visual BF. For all subjects, the RMS of (1)
trunk acceleration and (2) COP displacement, were lower in all conditions with eyes open
(linear, sigmoid, and off mode of visual BF) than with eyes closed (linear, sigmoid, and off
mode of audio BF). Table 1 shows the average absolute values of all parameters analyzed in
the 6 different conditions tested.
Postural Strategies
As Table 1 shows, the TIB mean activity was significantly greater during trials with BF
115
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
Table 1 – Average Values of The Parameters Analyzed in the Six Conditions Tested
Visual
Audio
Parameter
Off
Linear
Sigmoid
Off
Linear
Sigmoid
RMS COP [mm]
5.67(±1.34)
6.30(±1.90)
6.22(±2.38)
12.31(±2.48)
12.27(±2.81)
10.96(±2.02)
2
RMS Acc [mm/s ]
90.8(±26.9)
64.4(±22.5)
67.6(±23.1)
139.5(±35.3)
138.8(±40.9
119.2(±29.6)
RMS Shear [N]
0.09(±0.03)
0.09(±0.02)
0.10(±0.06)
0.19(±0.04)
0.21(±0.03)
0.19(±0.03)
AP Acc-COP correlation
0.62(±0.13)
0.50(±0.20)
0.47(±0.22)
0.68 (±0.12)
0.75(±0.16)
0.71(±0.13)
TIB mean activity
0.52(±0.62)
0.99(±1.25)
1.00(±1.08)
1.02(±1.19)
2.17(±1.50)
1.82(±2.30)
TIB-GAS co-contraction
0.09(±0.04)
0.13(±0.07)
0.12(±0.05)
0.09(±0.05)
0.06(±0.04)
0.10(±0.06)
(p<0.01) than in trials without BF. The EMG co-contraction between TIB-GAS did not
significantly change with BF.
The correlation between trunk acceleration and COP displacement in the AP direction
was significantly larger in trials with audio BF than in trials with visual BF (p<0.01, see Table
1). However, the correlation between trunk acceleration and COP displacement was not
significantly different between the two reference conditions (eyes open and eyes closed). The
shear forces were no different for the different BF modalities or coding but were significantly
larger for all the eyes closed conditions (none, linear audio, sigmoid audio BF) than for all the
eyes open conditions (none, linear visual, sigmoid visual BF).
Although linear visual BF and audio sigmoid BF both decreased trunk acceleration RMS
(in percentages 28.24±4.79 and 14.38±2.22, respectively), linear visual BF increased COP RMS
whereas audio sigmoid decreased COP RMS (11.2±6.6 and -8.59±3.29, respectively).
A. Acceleration in Anterior-Posterior Direction, Raw Data from a Representative Subject
mm/s2
50
Without ABF
0
With Linear ABF
50
10
15
20
25
30
35
40
45
50
55
With Sigmoid ABF
Time [s]
mm/s2
20
Without VBF
0
With Linear VBF
20
With Sigmoid VBF
10
15
20
25
30
35
40
45
50
55
Time [s]
B. COP Displacement in Anterior-Posterior Direction, Raw Data from a Representative Subject
50
mm
Without ABF
0
With Linear ABF
50
10
15
20
25
30
35
40
45
50
55
With Sigmoid ABF
Time [s]
50
mm
Without VBF
0
50
With Linear VBF
With Sigmoid VBF
10
15
20
25
30
35
40
45
50
55
Time [s]
Figure 4 – Panel A and panel B show trunk acceleration and COP displacement,
respectively, from a representative subject during all conditions tested.
116
Chapter 7
Discussion
The results reported in this paper showed how both visual and audio BF of acceleration
sensed at the trunk level reduced postural sway during upright stance. This sway reduction
with audio and visual BF is consistent with the sway reduction reported previously for
visual BF [2], for tactile BF [11;12], and for other types of audio BF [26;27]. In this study, two
different ways of coding trunk acceleration into BF presentation (linear and sigmoid) were
also tested.
The results reported in this paper show that sigmoid coding for audio BF and linear
coding for visual BF were the most effective to reduce postural sway in stance. Our results that
a different BF coding induces a different extent of sway reduction suggest that customized
BF coding for different modalities of BF will make BF information optimally usable. A more
sophisticated and accurate exploration of the possible coding between postural sway and BF
may result in an even larger sway reduction. For example, Rougier and colleagues found that
by adding a delay (>600 ms) and increasing the gain in the BF loop optimized the effects of
visual BF on postural stability [28].
Differences between how the nervous system naturally processes audio and visual inputs
for detecting postural sway may explain why different coding of BF are needed. Sigmoid
coding may be the best for audio BF because subjects can easily detect velocity of sway away
from initial posture by the rate of change (velocity) of pitch and volume. Coding feedback
with a sigmoid function results in very small changes in BF near the baseline, upright posture
with an increasing rate of change of BF as the subject leans toward their limits of stability. Jeka
et al. [29] suggest that velocity feedback from somatosensory and vestibular inputs is critical
for control of postural stability. Also, allowing a small area with no BF information of sway
near upright, as in the sigmoid coding, has the advantage of driving the subjects’ attention
to the BF only when it was needed. Some models of postural control suggest that natural
postural control includes a passive sway area without postural corrections until a threshold
is reached, when automatic postural adjustments are triggered [30].
Linear coding may be the best of visual BF because it depends upon subjects detecting
the difference in position of a visual signal (in this case, a star) relative to the position of a
target representing the initial, upright postural goal (in this case, an ellipse). This detection
of error in body versus target position for visual BF, and the relatively slow reaction times
elicited from visual inputs compared to auditory inputs [31], may be why a linear coding for
117
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
visual BF was optimal.
As expected, all the eyes closed conditions resulted in larger COP excursions, larger
trunk acceleration, and larger shear forces, consistent with other studies [19]. Because the
availability of a stable visual surround in the periphery has such a large effect on postural
sway, we normalized the effects of visual and audio BF to separate reference conditions with
eyes open, and eyes closed, respectively. Our previous study showed that audio BF has a very
limited effect when healthy subjects or vestibular loss subjects are standing on a firm surface
with eyes open, probably because of a ceiling effect, and because subjects are reluctant to
switch dependence from preferred sensory reference frames to novel sensory input for posture
[32]. However, the amount that subjects use auditory BF to reduce postural sway depends
upon how much it is needed based on the sensory context and the extent of their sensory
pathology with the maximum effect when vestibular loss subjects stand on a compliant
surface with eyes closed [32]. Thus, the optimal sensory mode for effective BF is likely to vary
under different sensory conditions, pathology and age. For example, trunk acceleration BF
information may be most effective for subjects who have lost otolith information whereas
COP BF may be more effective for subjects who have lost sensitivity to pressure under their
feet due to pathology.
While visual linear BF and sigmoid audio BF had the largest impact on postural sway,
each mode of BF appeared to facilitate a different type of postural sway movement strategy.
The visual BF mainly reduced trunk acceleration, whereas the audio BF mainly reduced COP.
These results, along with the greater correlation between COP and trunk acceleration that
was found in the audio BF condition, suggest that the two BF presentations induce different
postural, kinematics strategies. In fact, an inverted pendulum model of postural sway is
consistent with the effects of sigmoid audio BF. In contrast, linear visual BF, resulted in an
increase in COP displacement with a decrease in trunk acceleration and a lower correlation
between COP and trunk acceleration which is consistent with a multi-segmental model of
body sway [33]. The necessity of using two different kinematics models to explain the change
in postural movement strategy associated with audio and visual BF suggests that visual BF
pushes the control of posture more toward a “hip strategy” (multi-segmental model), and the
audio BF pushes the control of posture more toward an “ankle strategy” (inverted pendulum
model; [34;35]).
Our results also suggest that the eyes open reference condition was associated with
a larger contribution of hip strategy than the eyes closed reference condition [29]. In the
visual BF reference (eyes open) condition, the correlation between trunk acceleration and
COP displacement was lower than in the audio BF reference trial (eyes closed). In the eyes
open reference condition for visual BF, the strategy used to control posture may have had a
higher contribution of hip strategy [36] to fix the distance in space between head and monitor
whereas, in the eyes closed, reference trials for audio BF, the strategy used may have had a
118
Chapter 7
higher contribution of ankle strategy such that movement of the head and ears correlated
with movement of the body COM.
An alternative explanation could be that visual and auditory BF may not induce a
different strategy for the control of posture but, perhaps, simply enhance the natural postural
strategy already used by the central nervous system in that particular condition (eyes open
and eyes closed). If this hypothesis is confirmed, it could be further speculated that BF
increases the reliability of other sensory information by adding redundancy and providing a
reference which increases the signal-to-noise ratio in the control of posture sensory feedback
loop [37;38]. In other words, the central nervous system may use the BF information not only
by itself, but also in combination with the other sensory information to increase the precision
of the estimation of the body posture.
Neither visual nor auditory BF appeared to reduce postural sway via a stiffening strategy
since there was no increase in co-activation of muscles around the ankle joints. This suggests
that the added sensory information about body sway enhanced the natural, direction specific,
automatic postural control strategies rather than superimposing a generalized stiffening. The
increase in background TIB EMG activity during use of BF would reduce the threshold when
this ankle dorsiflexor would be recruited to resist backward and backward-lateral body sway
but was not associated with a change in background COP position in our subjects [39].
In conclusion, this study showed how reduction of postural sway in stance using BF
depends on the modality and coding of the BF of trunk acceleration. Linear visual BF and
sigmoid audio BF induced the largest reduction in postural sway although via different
postural kinematics strategies.
119
Effects of Linear vs Sigmoid Codingof Visual or Audio Biofeedback for the Control of Upright Stance
Bibliography
[1]
M. S. Schwartz and F. Andrasik, “Biofeedback in Neuromuscular Reeducation and Gait Training,” in
Biofeedback - Biofeedback: a Practitioners’s Guide, pp. 485-513, New York: Guilford Press, 2003.
[2]
F. Hlavacka and V. Litvinenkova, “First derivative of the stabilogram and posture control in visual feed-back
conditions in man,” Agressologie, vol. 14, no. Spec No C, pp. 45-49, Sept.1973.
[3]
M. C. Dault, L. Yardley, and J. S. Frank, “Does articulation contribute to modifications of postural control
during dual-task paradigms?,” Brain Res Cogn Brain Res, vol. 16, no. 3, pp. 434-440, May2003.
[4]
G. Wu, “Real-time feedback of body center of gravity for postural training of elderly patients with peripheral
neuropathy,” IEEE Trans Rehabil Eng, vol. 5, no. 4, pp. 399-402, Dec.1997.
[5]
A. Shumway-Cook, D. Anson, and S. Haller, “Postural sway biofeedback: its effect on reestablishing stance
stability in hemiplegic patients,” Arch Phys Med Rehabil, vol. 69, no. 6, pp. 395-400, June1988.
[6]
R. Hamman, N. S. Longridge, I. Mekjavic, and J. Dickinson, “Effect of age and training schedules on balance
improvement exercises using visual biofeedback,” J Otolaryngol, vol. 24, no. 4, pp. 221-229, Aug.1995.
[7]
S. E. Sihvonen, S. Sipila, and P. A. Era, “Changes in postural balance in frail elderly women during a 4-week
visual feedback training: a randomized controlled trial,” Gerontology, vol. 50, no. 2, pp. 87-95, Mar.2004.
[8]
C. J. Winstein, E. R. Gardner, D. R. McNeal, P. S. Barto, and D. E. Nicholson, “Standing balance training: effect
on balance and locomotion in hemiparetic adults,” Arch Phys Med Rehabil, vol. 70, no. 10, pp. 755-762,
Oct.1989.
[9]
L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans Biomed Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[10] J. Hegeman, F. Honegger, M. Kupper, and J. H. Allum, “The balance control of bilateral peripheral vestibular
loss subjects and its improvement with auditory prosthetic feedback,” J Vestib Res, vol. 15, no. 2, pp. 109117, 2005.
[11] C. Wall, III, D. M. Merfeld, S. D. Rauch, and F. O. Black, “Vestibular prostheses: the engineering and biomedical
issues,” J Vestib Res, vol. 12, no. 2-3, pp. 95-113, 2002.
[12] M. Tyler, Y. Danilov, and Bach-y-Rita P, “Closing an open-loop control system: vestibular substitution through
the tongue,” J Integr Neurosci, vol. 2, no. 2, pp. 159-164, Dec.2003.
[13] F. B. Horak, L. M. Nashner, and H. C. Diener, “Postural strategies associated with somatosensory and vestibular
loss,” Exp Brain Res, vol. 82, no. 1, pp. 167-177, 1990.
[14] L. M. Nashner, C. L. Shupert, and F. B. Horak, “Head-trunk movement coordination in the standing posture,”
Prog Brain Res., vol. 76, pp. 243-251, 1988.
[15] F. B. Horak, C. L. Shupert, and A. Mirka, “Components of postural dyscontrol in the elderly: a review,” Neurobiol
Aging, vol. 10, no. 6, pp. 727-738, Nov.1989.
[16] R. E. Mayagoitia, J. C. Lotters, P. H. Veltink, and H. Hermens, “Standing balance evaluation using a triaxial
accelerometer,” Gait Posture, vol. 16, no. 1, pp. 55-59, Aug.2002.
[17] L. Yardley, M. Gardner, A. Leadbetter, and N. Lavie, “Effect of articulatory and mental tasks on postural
control,” Neuroreport, vol. 10, no. 2, pp. 215-219, Feb.1999.
[18] S. A. Raper and R. W. Soames, “The influence of stationary auditory fields on postural sway behaviour in
man,” Eur J Appl Physiol Occup Physiol, vol. 63, no. 5, pp. 363-367, 1991.
120
Chapter 7
[19] L. Chiari, L. Rocchi, and A. Cappello, “Stabilometric parameters are affected by anthropometry and foot
placement,” Clin Biomech (Bristol. , Avon. ), vol. 17, no. 9-10, pp. 666-677, Nov.2002.
[20] T. E. Prieto, J. B. Myklebust, R. G. Hoffmann, E. G. Lovett, and B. M. Myklebust, “Measures of postural steadiness:
differences between healthy young and elderly adults,” IEEE Trans Biomed Eng, vol. 43, no. 9, pp. 956-966,
Sept.1996.
[21] L. Rocchi, L. Chiari, and A. Cappello, “Feature selection of stabilometric parameters based on principal
component analysis,” Med Biol Eng Comput, vol. 42, no. 1, pp. 71-79, Jan.2004.
[22] C. Maurer and R. J. Peterka, “A new interpretation of spontaneous sway measures based on a simple model
of human postural control,” J Neurophysiol, vol. 93, no. 1, pp. 189-200, Jan.2005.
[23] F. B. Horak and J.M. Macpherson, “Postural equilibrium and orientation,” in Handbook of Physiology. Rowell
RB and Shepherd J.T., Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[24] S. J. Olney and D. A. Winter, “Predictions of knee and ankle moments of force in walking from EMG and
kinematic data,” J Biomech, vol. 18, no. 1, pp. 9-20, 1985.
[25] R. J. Peterka and F. O. Black, “Age-related changes in human posture control: sensory organization tests,” J
Vestib Res, vol. 1, no. 1, pp. 73-85, 1990.
[26] A. M. Wong, M. Y. Lee, J. K. Kuo, and F. T. Tang, “The development and clinical evaluation of a standing
biofeedback trainer,” J Rehabil Res Dev, vol. 34, no. 3, pp. 322-327, July1997.
[27] M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback improves balance in patients with bilateral vestibular
loss,” Arch Phys Med Rehabil, vol. 86, no. 7, pp. 1401-1403, July2005.
[28] P. Rougier, “Compatibility of postural behavior induced by two aspects of visual feedback: time delay and
scale display,” Exp Brain Res, May2005.
[29] J. Jeka, T. Kiemel, R. Creath, F. Horak, and R. Peterka, “Controlling human upright posture: velocity information
is more accurate than position or acceleration,” J Neurophysiol, vol. 92, no. 4, pp. 2368-2379, Oct.2004.
[30] D. A. Winter, A. E. Patla, F. Prince, M. Ishac, and K. Gielo-Perczak, “Stiffness control of balance in quiet standing,”
J Neurophysiol, vol. 80, no. 3, pp. 1211-1221, Sept.1998.
[31] R. A. Schmidt, Motor control and leaning: a behaviral emphasis. Champaign IL: Human Kinetic Pub, 1982.
[32] M. Dozza, L. Chiari, and F. B. Horak, “Auditory Biofeedback Substitutes for Loss of Sensory Information in
Maintaining Stance,” Exp Brain Res, 2006.
[33] A. D. Kuo, “An optimal control model for analyzing human postural balance,” IEEE Trans Biomed Eng, vol. 42,
no. 1, pp. 87-101, Jan.1995.
[34] F. B. Horak and L. M. Nashner, “Central programming of postural movements: adaptation to altered supportsurface configurations,” J Neurophysiol, vol. 55, no. 6, pp. 1369-1381, June1986.
[35] R. Creath, T. Kiemel, F. Horak, R. Peterka, and J. Jeka, “A unified view of quiet and perturbed stance:
simultaneous co-existing excitable modes,” Neurosci Lett, vol. 377, no. 2, pp. 75-80, Mar.2005.
[36] S. M. Freitas, M. Duarte, and M. L. Latash, “Two kinematic synergies in voluntary whole-body movements
during standing,” J Neurophysiol, Nov.2005.
[37] A. D. Kuo, “An optimal state estimation model of sensory integration in human postural balance,” J Neural
Eng, vol. 2, no. 3, p. S235-S249, Sept.2005.
[38] R. A. Speers, A. D. Kuo, and F. B. Horak, “Contributions of altered sensation and feedback responses to changes
in coordination of postural control due to aging,” Gait Posture, vol. 16, no. 1, pp. 20-30, Aug.2002.
[39] S. M. Henry, J. Fung, and F. B. Horak, “EMG responses to maintain stance during multidirectional surface
translations,” J Neurophysiol, vol. 80, no. 4, pp. 1939-1950, Oct.1998.
121
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
122
Chapter
Chapter 8
Postural Responses Elicited by
Auditory-Biofeedback of Center of
Pressure during Perturbed Stance
Most of the content of this chapter will be submitted as: M. Dozza, L. Chiari, R.J. Peterka, C. Wall III, and F.B. Horak,
“Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance,” to Human
Movement Science.
123
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
124
Chapter 8
Abstract
Biofeedback is known to improve postural control by augmenting movement information.
However, the relation between amount of biofeedback information and postural control
improvements is still unknown. A few biofeedback-based products are now on the shelf
and promise to be effective for motor rehabilitation. However, the interaction between
spontaneous motor learning and biofeedback effect, which is the basis for the usefulness of
biofeedback in rehabilitation, is still unknown.
In this study, an audio-biofeedback system, providing different amounts of movement
information, was used to improve subjects’ performance during repetition of perturbed
stance.
Higher amount of audio-biofeedback information resulted in higher postural stability in
the beginning of the experiment. However, overtime, motor learning normalized the effects
of the different amount of audio-biofeedback information. Nevertheless, motor learning did
not neutralize the effect of audio-biofeedback at low frequencies (<0.2 Hz) of sway. Analysis
of postural responses transfer functions verified that audio-biofeedback affected prevalently
the low frequencies of sway (<0.4 Hz) whereas motor learning affected prevalently the high
frequencies of sway (>0.4Hz).
125
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
Introduction
The concept of biofeedback is well known since the 50’s [1]. However, only in the last few
years, the interest on biofeedback systems for postural control has renewed partially due to
the advance in technology. This renewed interest is evidenced by several recent publications
showing the efficacy of biofeedback in improving motor performances [2-5].
Despite these publications increase our knowledge about biofeedback, many questions
are still open about 1) biofeedback design, 2) biofeedback experimental protocols to be used
for rehabilitation, and 3) the mechanisms by which biofeedback system may induce postural
improvements and retention of motor performance.
The first challenge in the development of a biofeedback device is its design [6]. The
design of a biofeedback system should optimize the efficacy of its three main parts: 1) the
sensor unit, which acquires the information to be fed back; 2) the elaboration unit, which
processes and converts this biological information into new information; and 3) the restitution
unit, which conveys this new information to the user. However, to improve the design of
the whole biofeedback device, it is relevant to determine the amount of information that is
actually needed by the user, and the amount of information that the user is able to handle.
To date, there are no studies reporting on this issue.
Another challenge in the development of biofeedback devices is the protocol design to
be used for the device validation [7]. In fact, the experimental protocols at this stage of the
development should be aimed at evaluating the interaction between motor improvements
due to biofeedback and the motor improvements due to other mechanisms such as
spontaneous learning. This distinction is fundamental to evaluate retention and transfer
of motor performance after exposure to biofeedback and, finally, biofeedback efficacy for
rehabilitation. To date, very few studies reported on this issue, which is well known to be a
crucial one for the evaluation of biofeedback devices [7;8].
Up to now, biofeedback efficacy was determined, in most of the published studies, by
looking at some general balance indicators such as center of pressure, trunk angular velocity,
and head tilt which were also the feedback variables (e.g. [4;5;9], respectively). However
different biofeedback designs can induce different postural response strategies [10]. As a
consequence, to evaluate a biofeedback system, it is necessary to record a high number of
variables so that, not just the performance, but also the mechanisms and postural strategy
126
Chapter 8
used to achieve a better performance can be evaluated. For this evaluation, biofeedback
devices can take advantage of systems already available and purposely developed for
analyzing postural responses such as the one designed by Perterka [11]. Such device is able
to quantify postural response at different frequencies of induced sway so that a further insight
on the mechanism of sensory reweighting taking place during the exposure to biofeedback
can be achieved [12].
In this study, the amount of biofeedback information necessary to stabilize subjects in
perturbed stance and the interaction between the effect of biofeedback and spontaneous
learning during the practice of this task have been evaluated. Analyses from Peterka’s system
verified that biofeedback and motor learning affect different frequency intervals of postural
responses.
127
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
Materials & Methods
Participants
Thirteen healthy subjects, age 33±7 yrs, height
175±10 cm, and weight 78±18 Kg, participated to
this study. All subjects responded to the following
inclusion criteria: 1) no hearing deficits, 2) no
history of traumas or surgeries to the muscularskeletal system, and 3) no history of orthopedic
or neurological diseases or disorders. All subjects
signed an informed consent before the experiment
took place. This informed consent was approved
by the OHSU Ethical Committee and guaranteed
the subjects’ rights according to the Declaration of
Helsinki (1964).
Accelerometers
Position
Sensors
Movable
Platform
Figure 1 – Experimental set-up.
Protocol
All participants stood on a rotating force-plate able to destabilize their posture in the
medial-lateral (ML) plane (Fig. 1). The force-plate rotated accordingly to a pseudorandom
function [13] with a 4-degree peak-to-peak amplitude over a frequency range of 0.017
to 2.2 Hz [11]. In each trial, subjects were exposed to three cycles of the pseudorandom
perturbation. Each cycle was 60.5 s long, so that the total length of each trial was 181.5 s. All
participants were asked to maintain balance while the force-plate rotated and to respond to
the information from an audio-biofeedback (ABF) when available. This ABF was able to inform
the participants about their ML center of pressure (ML-COP) displacement according to four
different ABF modalities with different extent of information about ML-COP displacement. The
ML-COP displacement was recorded by the rotating force-plate. Two bi-axial accelerometers
(Analog Device ADXL202) were mounted on the subjects at C7 and L5 and were oriented so
that they could sense acceleration along the subjects’ anterior-posterior and ML direction. In
addition shoulder and hip position in the ML plane were recorded via two potentiometers.
Each subject was tested during three blocks of five randomized conditions. Four out of the five
conditions corresponded to the four different modalities of ABF, whereas the fifth condition
corresponded to a control condition where the subjects were not provided with any ABF.
128
Chapter 8
ABF modalities
The four ABF modalities differed in the amount of ML-COP information which was fed
back. Specifically, in modality 1, both the direction and the magnitude (full information) of
the ML-COP displacement were fed back to the subjects. In modalities 2 and 3, only direction
and only magnitude, respectively, was fed back to the subjects. Finally, in modality 4, the
ABF was limited to an alarm signal which informed the participants whenever their ML-COP
was exceeding a reference threshold (RT) in either the left or right direction. This RT was
determined for each subject based on their ML-COP displacement. This RT corresponded to
1 standard deviation of the ML-COP displacement recorded during the 10 seconds before
each trial. In all 4 modalities the ABF was provided only when the subject was exceeding this
threshold.
The ABF sound consisted of a 400-Hz sine wave modulated in volume so that changes
in volume could provide the information about the above-mentioned ML-COP displacement.
When subjects were inside the RT, the ABF volume was constant at 20 dB. The relations
between ABF volume and ML-COP displacement in the 4 different modalities are shown
in Figure 2. In particular, the algorithm controlling the relation ML-COP/volume in the first
modality (Fig. 2A) is the same described in Chiari et al., 2005 [14]. Briefly, when the subjects
move left/right: 1) the sound in the left/right earphone increases (20 to 50 dB) according to a
sigmoid function, and 2) the LR balance changes according to an exponential function so that
the sound in the earphone right/left earphone decreases (20 to 0 dB). In this way, both the
information about the direction and magnitude of the ML-COP displacement were provided
to the subjects. In the second modality, Fig. 2B, the volume of the sound was always the same
in both earphones and increased according to the same sigmoid function as used in the
first modality depending only on the magnitude of the ML-COP displacement. In the third
modality, Fig. 2C, the ABF volume changed according to a step function so that: 1) when the
subjects exceeded the RT in the left direction, the volume suddenly increased (0 to 50 dB) in
the left earphone and decreased (20 to 0 dB) in the right earphone; and 2) when the subjects
exceeded the RT in the right direction, the volume suddenly increased (0 to 50 dB) in the
right earphone and decreased (20 to 0 dB) in the left earphone. Thus, in this modality, the
only direction of the ML-COP displacement was provided to the subjects. Finally, in the fourth
modality, Fig. 2D, as soon as the subjects exceeded the RT the volume in both earphones
increased (20 to 50 dB) accordingly to a step function. Thus, the only information provided
to the subjects was if their ML-COP was inside or outside the RT. During the experiment, the
participants were asked to pay attention to the sound and to try to minimize its volume
which, lately, implied reducing their ML-COP displacement.
Data collection and analysis
For each trial, the COP displacement, the acceleration at L5 and C7, and the position
129
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
A.
B.
Volume
Volume
RT
Right Ear
Left Ear
RT
Right & Left Ear
ML COP Displacement
C.
Volume
ML COP Displacement
D.
Volume
RT
Right Ear
Left Ear
RT
Right & Left Ear
ML COP Displacement
ML COP Displacement
Figure 2 – ABF modalities – A: ABF coding both the magnitude and direction (full information)
of COP displacement. B: ABF coding only the magnitude of COP displacement. C: ABF coding only
direction of COP displacement. D: ABF coding only for exceeding the RF.
of hip and shoulder were recorded in the ML plane with a 100-Hz sample rate (Fig 1). From
ML-COP displacement and acceleration data the SDs were calculated. In addition, from the
position of hip and shoulder, the transfer function characterizing the subjects’ postural sway
responses (PTF) was calculated according to Peterka [11]. Briefly, from the hip and shoulder
position and anthropometry, the center of mass (COM) body sway angle with respect to
earth vertical was estimated. Then, the COM body sway angle and the measured rotation of
the force-plate were used to calculate the power spectra for each cycle of each trial. Finally,
the power spectra were averaged across the cycles to obtain a transfer function describing
the postural responses to force-plate rotation, in terms of gain and phase (at 16 frequencies
evenly spaced in the logarithmic frequency interval 0.016-2.2 Hz). Correlation analyses
were performed to verify the relation among COP, L5 acceleration, and C7 acceleration in
the ML plane. One-way ANOVA with Newman-Keuls multiple-comparison test was used to
verify significant difference (p<0.05) between trials with and without ABF (effect of the ABF
modality). Two-tail, paired T-test was used to verify significant difference (p<0.05) between
trials in the first and last block (effect of motor learning). Bonferroni correction was applied
in case of multiple comparisons.
130
Chapter 8
Results
ML COP and Accelerations
Effect of ABF on ML COP and Accelerations
When using ABF, all subjects exhibited a smaller sway compared to the control condition
in all three blocks of trials (Figure 3). All four ABF modalities significantly reduced ML COP
displacement and ML acceleration at L5 (Figure 4A-B). ML acceleration at C7 increased for
most of the subjects using ABF. However, this last result was not supported by statistical
significance. When averaged overtime, the effect of all ABF modalities on ML COP displacement
and accelerations was similar (Figure 4).
Effect of learning on ML COP and Accelerations
[degree]
[degree]
[degree]
[cm/s2]
[cm/s2]
[cm]
The amount of sway reduction caused by the four ABF modalities changed overtime.
Specifically, in the first block of trials, amount of sway (in terms of ML COP and acceleration
Without ABF
at L5; Figure 5A-B) was inversely
With Full-Information ABF
10 A. COP
proportional to the extent of information
0
coded by the ABF. In fact, full-information
-10
0
40
80
120
160
Time [s]
ABF resulted in the smallest amount of
B. L5 Acceleration
10
sway; alarm ABF resulted in the largest
0
-10
sway with ABF; and direction and
0
40
80
120
160
Time [s]
magnitude ABF resulted in a similar
C. C7 Acceleration
10
0
amount sway intermediate between
-10
0
40
80
120
160
the other two modalities. In the second
Time [s]
5
D. Surface tilt
and third block, the amount of sway with
0
the different ABF modalities was similar,
-5
0
40
80
120
160
Time [s]
even if the full-information ABF resulted
E. Ankle angle
5
in a slightly smaller sway compared
0
to the other modalities. Vice versa, in
-5
0
40
80
120
160
Time [s]
the first block, ML acceleration at C7
5
F. Hip angle
increased proportionally to the extent of
0
-5
information coded by the ABF. However,
0
40
80
120
160
Time [s]
in the second and third block ML
Figure 3 – Raw data from one representative subjects
acceleration at C7 was not significantly
in conditions 1 and 5 (i.e. with full-information ABF and
without ABF) from 2 trials performed in the second block of
different in all conditions tested.
131
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
Effect of ABF Modality on SD of the COP and L5 Acceleration
A. COP SD [mm]
11
Direction & Magnitude
Direction & Magnitude
21
Magnitude
9
B. L5 Acceleration SD [mm/s2]
Magnitude
Direction
Direction
Alarm
Alarm
no BF
20
no BF
19
7
18
17
5
*
*
*
*
*
*
*
*
Figure 4 – Averaged SD of A: center of pressure, and B: acceleration at L5 level in all five conditions
tested. Asterisks indicate significant difference (p<0.05) from control condition.
Effect of ABF Modalities on SD of COP and L5 Acceleration Overtime
A. COP SD
B. L5 Acceleration SD
23
Without ABF
Without ABF
Alarm
10
Alarm
Magnitude
Magnitude
Direction
Direction
Direction & M
Magnitude
21
[mm/s2]
[mm]
Direction & M
Magnitude
9
19
8
17
7
Block 1
Block 2
Block 3
Block 1
Block 2
Block 3
Figure 5 – Standard deviations of A: center of pressure and B: acceleration at L5 level overtime in all conditions
tested.
132
Chapter 8
Table 1 – Correlation coefficients (r) among center-of-pressure and accelerations.
Without ABF
With ABF
COP-Acc.L5
COP-Acc.C7
Acc.L5-Acc.C7
COP-Acc.L5
COP-Acc.C7
Acc.L5-Acc.C7
1st Block
0.89 (±0.04)
0.84 (±0.09)
0.77 (±0.12)
0.84 (±0.06)
0.58 (±0.21)
0.48 (±0.18)
2nd Block
0.90 (±0.03)
0.77 (±0.13)
0.69 (±0.15)
0.84 (±0.05)
0.71 (±0.12)
0.56 (±0.16)
rd Block
0.88 (±0.06)
0.77 (±0.16)
0.70 (±0.17)
0.84 (±0.06)
0.69 (±0.15)
0.56 (±0.20)
3
Effect of ABF and learning on the correlations between ML COP and Accelerations
Correlation among ML COP and acceleration increased when comparing overtime the
full-information ABF and the control condition. In the first block of trials, the correlations in
the ML plane between 1) COP and C7 acceleration, and between 2) L5 acceleration and C7
acceleration were lower in the full-information ABF than in the control condition (Table 1).
However, from the second block of trials, the same correlations in the ML plane between
1) COP and C7 acceleration, and between 2) L5 acceleration and C7 acceleration, increased
in the full-information ABF condition and decreased in the control condition ( Table 1).
Correlation between COP and L5 in the ML plane was high in both the full-information ABF
condition and in the control condition and did not significantly change overtime (Table 1).
Finally, correlations in the ML plane among COP, L5 acceleration and C7 acceleration did not
significantly change between the 2nd and the 3rd block of trials.
Postural Response Transfer Function
Effect of ABF on PTF gain
ABF reduced the gain of the PTF especially at low frequencies. Figure 6A shows data
from a representative subject, in the three blocks of trials, comparing the full-information
ABF and control condition. Further, the effect of ABF (averaged across subjects) in the first
and last block of trials are reported in Figure 7A-B. In the first block full-information ABF
significantly (p<0.05) decreased the PTF gain at the very low frequency (0.02 Hz) and in a
narrow interval around 1 Hz (Figure 7A). In the third block, full-information ABF significantly
(p<0.05) decreased the PTF gain in the wide interval 0.02-0.2 Hz (Figure 7B). Figure 7C
compares the effects, in terms of PTF gain reduction, occurred in the first and third block.
Specifically, in the first block the largest gain reduction occurred around 0.8 Hz. In the third
block the largest gain reduction occurred at low frequencies (< 0.2 Hz).
Effect of learning on PTF gain
Subjects reduced the gain of the PTF overtime in all conditions tested. Figure 6B shows
data from a representative subject, in all conditions tested, comparing the first and third block
of trials. Further, the effect of time (averaged across subjects) in the full-information ABF
and control condition are reported in Figure 8A-B. Both the full-information ABF and control
condition significantly (p<0.05) decreased the PTF gain overtime at the very low frequency
133
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
(0.02) and in the interval from 0.2 to
1.1 Hz (Figure 8A-B).
Effect of ABF and Learning on the Gain of the Postural Reponse Transfer Function
A. Trials without (black) and with ABF (gray) in the thee blocks
Block 1
Block 2
0
Block 3
10
-1
Gain
Gain
Gain
10
-1
-1
10
10
10
0
-3
1
10
10
-2
10
0
-3
10
-2
1
10
10
Frequency [Hz]
0
-3
10
10
1
10
10
10
Frequency [Hz]
Frequency [Hz]
range of frequencies both in the fullinformation ABF and control condition.
By plotting the difference between the
PTF in the first and third block, it was
possible to highlight a peak of gain
reduction (due to the reduction of
gain in the narrow range of frequency)
for each subject. Each subject showed
the peak of gain reduction at slightly
different frequenc y but always
comprehended between 0.2 Hz and
0.9 Hz. This peak was presented both
B. First (black) and third (gray) block in the five conditions
Direction
-1
-2
10
-1
10
0
10
1
Gain
10
-1
10
0
10
1
-2
10
10
Frequency [Hz]
-1
10
10
10
-2
10
Frequency [Hz]
10
-1
-1
10
Withuot ABF
0
10
10
-1
10
Alarm
0
0
Gain
10
Gain
10
Magnitude
0
Gain
0
Gain
Direction & Maginitude
-1
10
0
10
1
10
Frequency [Hz]
-2
-1
10
10
0
1
10
10
-2
10
-1
10
0
10
1
10
Frequency [Hz]
Frequency [Hz]
Figure 6 – Gains of the PTF (i.e. ratio between body sway
amplitude and stimulus amplitude) from one representative
subject. Panel A represent the effect of using full-information
ABF (Condition 1) compared to no ABF (Condition 5) in the three
blocks of trials. Panel B represents the learning effect i.e. the
comparison between trials in the first and last block.
Effect of ABF: Difference between with and without ABF
A. First Block
B. Third Block
0
A. With ABF
0
10
0
10
Gain
-1
10
-1
10
-1
10
Without ABF
-2
10
-1
First Block
Third Block
0
1
10
10
10
-2
10
10
-1
0
10
10
1
10
-2
10
10
Frequency [Hz]
0.4
Significant (p<0.05) learning without ABF
-1
-2
10
-2
10
-1
0
1
10
10
Significant (p<0.05) learning with ABF
Not significant learning with ABF
10
10
Frequency [Hz]
Not significant ABF effect in the third block
0
10
1
10
Gain Reduction
Gain Reduction
1
10
Not significant ABF effect in the first block
0.1
Figure 7 – Effect of full-information ABF on PTF
gain (i.e. ratio between body sway amplitude and
stimulus amplitude) in the first (panel A) and last
(panel B) block. The effects of full-information ABF in
the first and third block are compared in panel C.
Not significant learning without ABF
0.2
0
-2
10
10
-1
0
10
1
10
Frequency [Hz]
Frequency [Hz]
-0.1
134
0
10
C. With vs Without ABF
Significant (p<0.05) ABF effect in the third block
-2
-1
Frequency [Hz]
Significant (p<0.05) ABF effect in the first block
10
Significant difference (p<0.05)
between the first and last block
-2
-2
C. First vs Third Block
0
Third Block
Significant difference (p<0.05)
between the first and last block
Significant difference (p<0.05)
between with and without ABF
Frequency [Hz]
0.2
-1
First Block
With ABF
-2
10
10
Without ABF
With ABF
Significant difference (p<0.05)
between with and without ABF
10
B. Without ABF
0
10
Gain
10
Gain
Learning: Difference between the First and Third Block
Gain
-2
10
The extent of gain reduction
overtime was the largest in a specific,
narrow interval of frequencies. In
particular, all subjects showed the
largest gain reduction in a narrow
0
0
10
-0.2
Figure 8 – Effect of learning on PTF gain (i.e.
ratio between body sway amplitude and stimulus
amplitude) with full-information ABF (Condition 1;
panel A) and without ABF (Condition 5; panel B). The
effects of learning with and without full-information
ABF are compared in panel C.
Chapter 8
Effect of ABF: Difference between with and without ABF
A. First Block
B. Third Block
100
0
Phase [deg]
Phase [deg]
100
-100
0
-100
Without ABF
Without ABF
With ABF
With ABF
Significant difference
(p<0.05) between with
and without ABF
-200
-2
10
10
-1
Significant difference
(p<0.05) between with
and without ABF
-200
0
10
1
-2
10
10
Frequency [Hz]
10
-1
0
10
1
10
Frequency [Hz]
Learning: Difference between the First and Third Block
C. With ABF
100
D. Without ABF
100
0
Phase [deg]
Phase [deg]
in the full-information ABF and in
the control condition. However, the
position and amplitude of the peak
was not always the same in all subjects
in between these two conditions.
The gain reduction, averaged across
subjects, is show in Figure 8C for
the full ABF information and control
condition. The two low peaks in Figure
8C are the consequence of averaging
the individuals gain reduction peaks.
In the full-information ABF condition,
the low peak was lower in amplitude
and frequency compared to the low
peak in the control condition (Figure
8C). However, this difference between
the low peaks was not verified in each
subject data.
-100
0
-100
First Block
First Block
Third Block
Third Block
Effect of ABF and learning on PTF phase
Significant difference
(p<0.05) between the
first and third block
-200
Significant difference
(p<0.05) between the
first and third block
-200
ABF and learning increased
10
10
10
10
10
10
10
10
PTF phase in different intervals of
Frequency [Hz]
Frequency [Hz]
frequency. Specifically, ABF significantly
Figure 9 – PTF phase changes due to full-information ABF in
the first (A) and third (B) block of trials. PTF phases changes
affected low frequencies (<0.4 Hz)
overtime with full-information ABF (C) and without ABF (D).
whereas learning significantly affected
high frequencies (>0.8 Hz). In addition, the amplitude of PTF phase increase due to ABF was
larger than the increase due to learning. Figure 9 shows the effect of ABF and learning on
PTF phase.
-2
-1
0
1
-2
-1
0
1
135
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
Discussion
During this experiment, two main factors concurred in decreasing subjects’ sway: 1)
use of ABF (i.e. augmented sensory information), and 2) repetition of the task (i.e. motor
learning). These two factors interacted during the experiment inducing a similar extent of
sway reduction. For this reason, they need to be considered combined for the interpretation
of the results.
ML COP and Accelerations
With all ABF modalities, subjects’ improved their balance even after motor learning
occurred (block 3). However, only in the first block of trials, significant differences between
the ABF modalities were evidenced by the subjects’ performance. In fact, in the first block,
the advantage of having ABF with a larger amount of ML-COP information resulted in better
performances. In the second block, the difference among the four modalities of ABF became
less evident from the subjects’ performance, and this difference, then, almost disappeared in
the third block. For this reason, the effects of all ABF modalities averaged across time (Figure
4) do look similar. Nevertheless, subjects’ exhibited smaller sway only when they received ABF,
even in the third block of trials where sway reduction induced by motor learning was the
maximum. This result is consistent with our previous results where we showed how this ABF
system decrease postural sway in normal and vestibular loss subjects [15].
Sway reduction induced by motor learning is evidenced by the subjects reducing sway
overtime in the control condition. However the extent to which motor learning was induced
by ABF or was spontaneous is open to debate. In fact, in other experiments, not involving
ABF [11], subjects did not show sway reduction overtime by simple practice of standing on
the same rotating force-plate. This finding supports the hypothesis that ABF favored motor
learning during the experiment resulting in motor retention during control trials.
Subjects performance improved overtime also in all ABF modalities. This result confirms
that some learning mechanism occurred during the trial. However, the extent to which
improvements in ABF condition were driven by motor learning or by an optimization of the
ability to use ABF is still questionable. In fact, a better performance in the task may have
been achieved by a more correct interpretation of ABF, which may have been developed
overtime by the subjects. Also, the result that, in the third block, subjects could achieve similar
performances independently from the modality of ABF, suggests that the subjects were able
136
Chapter 8
to implement faster and more accurate corrections in response to the ABF. Thus, subjects may
have completed the partial information from the different ABF modalities with natural sensory
information. In this case, improvement of postural performance overtime using ABF may be
the result of an improved integration among sensory information and between sensory and
artificial information.
In addition, the task tested in this study is not critical for healthy adults – as the subjects
participating in this study. Thus, the similar performance, achieved by the subjects overtime,
independently from the ABF modality used, may depend on a ceiling effect. In other words,
overtime subjects may have been able to optimize their ability of maintaining stance on the
rotating surface to the point that further improvements were not possible by simply adding
ML COP information.
Movements at the hip level were restricted at the beginning of the experiment and
became less restricted overtime in the control condition. In fact, in the control condition,
correlation in the ML plane between COP and C7 acceleration and L5 acceleration and C7
acceleration decreased overtime. When using ABF, subjects presented from the very beginning
a low correlation in the ML plane between COP and C7 acceleration and L5 acceleration and
C7 acceleration which then increased with the optimization of postural responses and the
consequent reduction of sway. Thus, using ABF, movements at hip level were not restricted
at the beginning and then became more restricted once the task became easier. However,
the best performance in all condition was achieved when movements at the hip level were
evident (third block).
Also, standard deviation of correlation factors increased overtime in all condition,
suggesting that subjects were not converging to a common strategy for the control of posture
but, instead, were taking advantage of personalized multi-segmental control to achieve best
performances. ABF favored from the beginning multi-segmental control of posture, somehow
anticipating what spontaneous control of posture may have developed overtime. Thus, ABF
may have favored spontaneous motor learning by inducing the subjects to gain confidence
with multi-segmental control of posture from the very beginning of the experiment.
Postural Responses Transfer Function
ABF and motor learning resulted in gain reduction in the PTF – a lower gain is indicator
of higher stability [11]. Once again the two factors (ABF and motor learning) causing gain
reduction acted contemporarily and with a similar extent on subjects’ gain. However, by
analyzing the PTF gain and phase changes at the different frequencies, it is possible, to
partially discriminate the effect of learning and ABF.
The effect of learning is evident from the gain reduction occurred at 0.02 Hz and in the
range 0.2-1 Hz overtime. This gain reduction was found to be similar for trials with and without
ABF, suggesting that subjects converged overtime to the same postural control mechanism in
137
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
both ABF and control condition. However, only in the first block, the instantaneous effect of
ABF regarded the same frequencies affected overtime by learning. In other words, the effect
of learning was similar to the effect of using ABF in the first block. Then, by the third block the
effect of ABF regarded only the frequencies below 0.2Hz. This result suggests that, overtime,
subject retained motor improvements in the range 0.2-1Hz but not at low frequencies (<0.2
Hz) where ABF continued to show an additional reduction effect on top of motor learning.
Since gain reduction was found to be an indicator of sensory reweighting from somatosensory
to vestibular [11;16], the gain reduction induced by ABF may be also the consequence of a
sensory reweighting which favored vestibular information over somatosensory information.
The effect of ABF on sway was also somehow similar to the effect of a vestibular
prosthesis based on tactile biofeedback [17] which was recently tested with the same
perturbation used in this study [12]. Using this prosthesis both control subjects and bilateral
vestibular loss subjects 1) reduced sway, 2) reduced PTF gain at low frequencies (<0.8 Hz), 3)
increased PTF gain at high frequencies (>0.8Hz), and 4) did not change PTF phase. Some of the
reasons why the results reported in this study differ from the ones reported by Perterka et al.
[12] may be due to: 1) the different design of the biofeedback devices used (ABF versus tactile
biofeedback), 2) the different direction of biofeedback information and platform perturbation
(ML in this study versus anterior posterior in Perterka’s study), and 3) the use of a back board
only in Peterka’s study which constrained the subject to move as an inverted pendulum.
Peterka et al. [12] suggest that the PTF gain reduction at low frequencies (which was found
also in this study) could be due to the limited bandwidth of the orientation information from
the biofeedback. However, another explanation could be that high frequency orientation
information had been filtered out by the intrinsic delay of the voluntary postural response
to ABF. In other word, high frequency gain reduction would inevitably require short time
responses which may not be compatible with the several-hundreds-millisecond dynamic
needed for the brain to receive the biofeedback information, elaborate it, and activate the
muscles to generate the postural response. In this case, practicing could improve the balance
prosthesis performance by making more automatic the postural responses to the orientation
information [18;19]; the gain reduction found up to 1.1 Hz in this study after practicing
supports this last speculation.
An unexpected result of this study was that subjects did not reduce the PTF gain at all
frequencies overtime but, instead, had a pretty narrow and specific range of frequencies that
they tended to reduce the most. This narrow range matches the range of frequencies where a
small peak, similar to a resonance peak in a second-order system, was also evident in the PTF.
The presence of such peaks in a transfer function normally determines more instability for
the system in the range of frequencies where the peak is. As a consequence, the reduction of
gain in a narrow range of frequencies matching the frequencies of the PTF gain peak seems
aimed at improving the system stability where it was more needed. In other words, the peak
of reduction showed by the subjects in some narrow range of frequencies may have been
138
Chapter 8
favored by the system being a priori more instable in that very range of frequencies.
Once motor learning occurred, PTF phase showed a clear difference between the effect
of ABF and learning. In fact, the effect of ABF and learning regarded two distinct different
intervals of frequency (Figure 9). In particular, ABF anticipated the phase delay at low
frequencies (<0.4 Hz) whereas learning anticipated the phase delay at high frequencies (>0.8
Hz). These results suggest that 1) postural responses to low-frequency perturbation were
faster when using ABF; and 2) postural responses to high-frequency perturbation became
faster with repetition of the task;
In conclusion, this study showed how motor learning and sensory augmentation concur
to sway reduction when humans are practicing a dynamic task, such as perturbed stance,
using an ABF system. Higher amount of ABF information resulted in higher postural stability
in the beginning of the experiment. However, overtime, motor learning normalized the effects
of the different ABFs. Nevertheless, motor learning did not neutralize the effect of ABF at low
frequencies (<0.2 Hz) of sway. With learning, subjects increased the variability of postural
control by using a multi-segmental strategy. With ABF, subjects used from the very beginning
a multi-segmental strategy that was then optimized overtime. PTF analysis highlighted some
differences among the mechanisms by which motor learning and ABF caused sway reduction
once motor learning occurred. In particular, motor learning favored PTF gain reduction in the
0.2-1 Hz interval and PTF phase increase above 0.8 Hz whereas ABF favored PTF gain reduction
for the frequencies below 0.2 Hz and increase of PTF phase below 0.4 Hz
139
Postural Responses Elicited by Auditory-Biofeedback of Center of Pressure during Perturbed Stance
Bibliography
[1]
M. Schwarz and Olson P.R., “A hystorical perspective on the field of biofeedback and applied
psychophysiology,” in Biofeedback: a practitioner’s guide, 3rd ed. M. Schwarz and F. Andrasik, Eds. New
York: Guilford, 2003, pp. 3-19.
[2]
C. Wall, III and E. Kentala, “Control of sway using vibrotactile feedback of body tilt in patients with moderate
and severe postural control deficits,” J. Vestib. Res., vol. 15, no. 5-6, pp. 313-325, 2005.
[3]
M. Dozza, L. Chiari, and F. B. Horak, “Audio-biofeedback improves balance in patients with bilateral vestibular
loss,” Arch. Phys. Med. Rehabil., vol. 86, no. 7, pp. 1401-1403, July2005.
[4]
M. Tyler, Y. Danilov, and Bach-y-Rita P, “Closing an open-loop control system: vestibular substitution through
the tongue,” J. Integr. Neurosci., vol. 2, no. 2, pp. 159-164, Dec.2003.
[5]
J. Hegeman, F. Honegger, M. Kupper, and J. H. Allum, “The balance control of bilateral peripheral vestibular
loss subjects and its improvement with auditory prosthetic feedback,” J. Vestib. Res, vol. 15, no. 2, pp. 109117, 2005.
[6]
H. Huang, S. L. Wolf, and J. He, “Recent developments in biofeedback for neuromotor rehabilitation,” J.
Neuroengineering. Rehabil., vol. 3:11., p. 11, June2006.
[7]
S. L. Wolf, “Electromyographic biofeedback applications to stroke patients. A critical review,” Phys. Ther.,
vol. 63, no. 9, pp. 1448-1459, Sept.1983.
[8]
S. T. Moore and M. H. Woollacott, “The use of biofeedback devices to improve postural stability,” Phys. Ther.
Practice, no. 2, pp. 1-19, 1993.
[9]
M. C. Dault, L. Yardley, and J. S. Frank, “Does articulation contribute to modifications of postural control
during dual-task paradigms?,” Brain Res Cogn Brain Res, vol. 16, no. 3, pp. 434-440, May2003.
[10] M. Dozza, L. Chiari, F. Hlavacka, A. Cappello, and F. Horak, “Effects of Linear versus Sigmoid Coding of Visual
or Audio Biofeedback for the Control of Upright Stance,” IEEE Trans. Neural Syst. Rehabil. Eng, 2006.
[11] R. J. Peterka, “Sensorimotor integration in human postural control,” J. Neurophysiol., vol. 88, no. 3, pp. 10971118, Sept.2002.
[12] R. J. Peterka, C. Wall, III, and E. Kentala, “Determining the effectiveness of a vibrotactile balance prosthesis,”
J. Vestib. Res., vol. 16, no. 1-2, pp. 45-56, 2006.
[13] W. D. T. Davies, System identification for self-adaptive control. London: Wiley-Interscience, 1970.
[14] L. Chiari, M. Dozza, A. Cappello, F. B. Horak, V. Macellari, and D. Giansanti, “Audio-biofeedback for balance
improvement: an accelerometry-based system,” IEEE Trans. Biomed. Eng, vol. 52, no. 12, pp. 2108-2111,
Dec.2005.
[15] M. Dozza, F. Horak, and Chiari L., “Auditory Biofeedback Substitutes for Loss of Sensory Information in
Maintaining Stance,” Exp. Brain Res, 2006.
[16] M. Cenciarini and R. J. Peterka, “Stimulus-dependent changes in the vestibular contribution to human
postural control,” J. Neurophysiol., vol. 95, no. 5, pp. 2733-2750, May2006.
[17] C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical sensors
using vibrotactile feedback of tilt,” IEEE Trans. Biomed. Eng, vol. 48, no. 10, pp. 1153-1161, Oct.2001.
[18] E. De Visser, R. P. Veth, H. W. Schreuder, J. Duysens, and T. Mulder, “Reorganization of gait after limb-saving
surgery of the lower limb,” Am. J. Phys. Med. Rehabil., vol. 82, no. 11, pp. 825-831, Nov.2003.
[19] M. C. Dault and J. S. Frank, “Does practice modify the relationship between postural control and the
execution of a secondary task in young and older individuals?,” Gerontology, vol. 50, no. 3, pp. 157-164,
May2004.
140
Chapter 8
141
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
142
Chapter
Chapter 9
Effect of Trunk-Tilt Tactile
Biofeedback on Tandem Gait in
Vestibular Loss Subjects
Most of the content of this chapter will be submitted as: M. Dozza, R.J. Peterka, C. Wall III, L. Chiari, and F.B. Horak,
“Effects of Practicing Tandem Gait with and without Vibrotactile Biofeedback in Subjects with Unilateral Vestibular Loss,”
to Experimental Brain Research.
143
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
144
Chapter 9
Abstract
Subjects with unilateral vestibular loss exhibit motor control impairments as shown
by body and limb deviation toward the affect side during gait. Biofeedback devices have
been showed to improve postural control, especially when sensory information is limited
by environmental conditions or pathologies such as unilateral vestibular loss. However, the
extent to which BF could improve motor performance or learning while practicing a dynamic
task such as narrow gait is still unknown. In this study 9 unilateral vestibular loss subjects
practiced narrow gait in 2 practice sessions with and without wearing a trunk-tilt biofeedback
device. The biofeedback device informed the subjects of their medial lateral angular tilt and
tilt velocity during gait via vibration of the abdomen. From motion analysis and tilt data, the
performance of the subjects practicing tandem gait were evaluated overtime and with and
without biofeedback.
By practicing tandem gait, subjects reduced their trunk-tilt, center of mass displacement,
variability of stepping, and frequency of stepping error. In both groups, use of biofeedback
consistently increased postural stability during tandem gait. Use of tactile biofeedback
consistently improved performance of unilateral vestibular loss subjects while they practiced
narrow gait. However, one session of practice with biofeedback did not result in conclusive
after-effects consistent with retention of motor performance without this additional
biofeedback. Tactile biofeedback acts similar to natural sensory feedback in improving
dynamic motor performance and not as a method to recalibrate motor performance to
improve function after short-term use.
145
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Introduction
Integration of vestibular, visual, and somatosensory information is fundamental to
maintain balance and to perform motor tasks [1]. When sensory information is missing, as
for example in subjects with unilateral vestibular loss (UVL), postural control is impaired and
subjects show an increased reliance on visual and somatosensory information [2]. UVL is often
a consequence of unilateral vestibular neurotomy to remove an acoustic neuroma [3]. After
the neuroma removal, subjects undergo a period during which the central nervous system
relearns how to cope with mismatching sensory information from vestibular, proprioceptive,
and visual senses [4]. Although subjects show improvement of balance after surgery, [4], most
UVL subjects, even years afterwards, continue to show balance and vestibular disorders such
as 1) inability to stand with eyes closed on a sway-referenced surface [5], 2) body and limb
deviation toward the affected side with eyes closed [2] and 4) difficulty balancing with eyes
closed a) on one foot, b) in tandem stance, and c) in tandem gait [6].
Sensory information can be augmented by using a biofeedback (BF) system [7]. BF
systems have been suggested to be beneficial when aimed to improve daily living tasks such
as gait [8]. However, most of the published studies to date have investigated the use of BF
systems during static or “quasi-static” tasks such as quiet or perturbed stance [9-13].
BF systems for postural control aim to encode some crucial kinematic or kinetic
information not normally accessible to subjects into information useful for nervous system
control of the task [7]. For example, during gait, information about trunk movement in the
medial-lateral plane is crucial for postural stability [14].
Visual, acoustic, and tactile BF systems have been used successfully to improve stance
balance in subjects lacking vestibular, visual, and somatosensory information [9], [15], [16],
respectively. However, use of visual and acoustic BF systems, could interfere with the ability
to deal with visual and acoustic information important for daily living. Thus, tactile BF may
be more suitable than visual or acoustic BF for providing additional feedback to improve
balance during daily living activities [17].
Improvements in specific motor tasks after practice with BF have been reported in many
studies [18]. However, practice of a specific motor task itself stimulates brain plasticity and
improves motor performances [2]. Thus, unless a control group is used to determine the
extent of spontaneous learning, the effects of BF on retention of motor performance remain
146
Chapter 9
inconclusive [19]. Knowing the extent to which BF practice facilitates retention of postural
performance improvements could help determine if BF intervention should be temporary
(used only during exercise sessions) or permanent (used as a prosthesis device).
In this study, the effect of augmented medial-lateral trunk tilt information via tactile BF
during repetition of a tandem gait task was assessed in subjects with UVL. Further, a crossover design in the experimental protocol was used to limit order effect when comparing the
short-term retention effects of practicing tandem gait with and without trunk tilt BF.
147
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Materials and Methods
Subjects
Table 1 – Details on the UVL Subjects Involved in this Study.
Subject ID
Age
Sex
Years Postsurgery
Pathology
Side Affected
Group
1
60
F
12
Ac. Neuroma
Right
1
Nine UVL subjects (5 males
2
26
M
10
Skull Injury
Left
1
and 4 females, age: 49±11yrs, height:
3
43
F
8
Ac. Neuroma
Left
1
4
46
F
8
Ac. Neuroma
Right
1
172±10cm, and weight: 89±21kg)
5
53
M
4
Ac. Neuroma
Right
1
participated in this experiment after
6
56
M
3
Ac. Neuroma
Left
2
signing an informed consent. This
7
63
F
n/a
Ac. Neuroma
Left
2
informed consent was approved
8
42
M
7
Ac. Neuroma
Right
2
9
49
M
n/a
Labyrinthitis
Right
2
by the academic, ethic committee
and guaranteed the subjec ts’
rights according to the Declaration of Helsinki. All subjects were free from orthopedic and
neurological diseases or disorders, except for the total vestibular loss on either the left (6
subjects) or the right (3 subjects) side. Subjects’ demographics and pathologies are shown
in Table 1.
As part of the cross-over experimental design, the subjects were divided into two groups
such that the differences between averaged ages, heights, and weights in the two groups
were not statistically significant (p < 0.05) when compared with a 2-tailed t test.
Apparatus
During the experiments, the subjects were asked to tandem-walk heel to toe, on a firm
surface while a commercial metronome was set to “beep” at 30 beats per minute (0.5 Hz). To
assure consistent cadence, subjects were asked to take one step for each beep. The subjects’
kinematics was acquired using a Motion Analysis system with 8 Falcon cameras. A symmetric
set of 20 markers was used (Figure 1). The markers were fixed above the eye, on the jaw joint,
and on the acromion, elbow, wrist, great trochanter, knee, malleolus, fifth metatarsal, and
hallux of each side of the subject. During all trials, the subjects were wearing a vibrotactile
BF system [17] constituted of a vest with 4 columns of tactors (3 tactors per column) and a
one-axis tilt sensor unit. The vest was placed around the trunk of the subject with an elastic
girdle so that 2 columns of tactors were in contact with the left side of the subject’s trunk
and the other 2 columns in contact with the right side of the subject’s trunk. The sensor unit
was aligned so that it could sense the subject’s medial-lateral (ML), trunk tilt. The sensor
148
Chapter 9
unit consisted of a rate gyroscope and
a linear accelerometer. A specially
developed algorithm combined
these two inputs to produce an
[mm]
1600
estimate of the subject’s orientation
to the vertical that was accurate
1200
to within 0.2 degrees. In particular,
the angular velocity sensed by the
800
gyroscope was high-pass filtered,
[mm]
integrated, and then summed to the
400
2000
low-pass filtered acceleration sensed
1500
0
by the accelerometer [20] . Before
1000
0
each experimental trial, the software
500
400
allowed the experimenter to “zero”
800 [mm] 0
the instrumentation while the subject
Figure 2 – 3D reconstruction of the experimental set-up.
Markers from Motion Analysis are represented as black
stood quietly in a vertical position. The
spheres. Trace of the center-of-mass, calculated from the
Motion Analysis data and the subject’s anthropometric
sensor unit was mounted on the right
measures, is also represented.
side of the subjects at L5 level using
a VelcroTM belt. This position was preferred because it is close to the center of mass (COM)
and minimally affected by artifacts such as breathing and heart beat. A computer (Macintosh
Powerbook G3) was used to activate the tactors on the vest depending on the subject’s ML,
trunk tilt detected by the sensor unit. The tactors on each side were activated in pairs using
a step-wise scheme depending on a combination of angular tilt and angular tilt velocity [21].
The lowest pair was activated when the sum of the measured tilt and one half of the measured
tilt velocity exceeded a 2 degree “dead-zone”, switching to the middle pair at 7 degrees and to
the highest when exceeded 12 degrees (Figure 2). During the experiment, all subjects wore
exactly the same type of polyester T-shirt so that the intensity of the vibration was as similar
as possible for each subject. Data were acquired with a 120-Hz sample frequency from the
tilt sensor and 60-Hz from the Motion Analysis system.
Procedure
Before starting the data collection, all UVL subjects learned how to perform tandem
gait safely and correctly during a 5- to 10-min-long training period. During this training, all
UVL subjects gradually learned how to take one step for each beat from the metronome
while keeping their eyes closed and arms crossed. All subjects, at first, were very skeptical
about their ability to tandem walk with eyes closed. However, after this very short training, all
subjects were able to successfully complete all trials. At the very beginning of the training
period, subjects had difficulty maintaining balance and made large lateral trunk movements.
Subjects attempted to compensate by using wider lateral foot placements during gait. This
149
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Figure 2 – Qualitative effect of trunk tilt on tactors activation. Circles on the left and right side of the human figure
indicate the tactors. Dark-filled circles indicate tactors on, empty circles indicate tactors off.
effect was controlled by monitoring the ML deviation of the foot placement across the steps
and providing verbal feedback to correct performance. Once subjects gained more confidence
with narrow stance, they tended to walk as fast as possible so that their own body inertia
helped to maintain balance. This effect was controlled by monitoring the actual frequency
of step.
Following training, each subject performed a test session of 30 trials of tandem walking
barefoot with eyes closed and arms crossed, taking one step for each beep of the metronome.
A second identical test session of the experiment was performed two weeks after the first
one. We refer to these two test sessions as “practice sessions” because we hypothesized that
motor learning would occur during each session and that results from the second session
would be influenced by practicing tandem gait in the first session.
In both sessions, the first 3 and the last 3 trials were performed with the tactile-BF device
turned off. A cross-over design was used for the 24 middle trials. Group 1 performed the
24 middle trials of the first session with the tactile-BF device turned off and the 24 middle
trials of the second session (two weeks later) with the device on. For Group 2 this order was
reversed.
Each walking trial was 2.5 meters long so that the subjects could take at least 5 complete
steps. Before the first session of the experiment, the subjects practiced the task for 5 to 10
minutes in order to get familiar with tandem walking. Subjects started practicing with eyes
open and without the metronome, then with eyes closed, and finally with eyes closed and
the metronome. Data collection started once the subjects understood the task and they
demonstrated that they were able to perform such a challenging task. At the beginning of the
practice period, all the subjects stated they would never be able to perform tandem walking
with eyes closed, however all of them actually could achieve this for a couple of meters after
the 10-minute practice period. During the experimental session, a safety spotter from our
laboratory walked on one side of the subjects to catch them in case they lost balance.
Data- and Statistical- Analysis
From the kinematics data and the anthropometric measures of each subject, the 3Dcoordinate of the COM during each trial was calculated according to [22-24]. Trunk tilt and
150
Chapter 9
COM data were synchronized via recording of trigger signals from Motion Analysis. Steps
were recognized from the position in time of the 6 markers on the feet. The first and the last
stance phases of stepping were neglected for the calculation of the following parameters.
The ML SD of the COM was calculated for each trial and used as an indicator of subjects’
ML postural stability. The standard deviation (SD) of ML tilt from the BF system sensor was
calculated for each trial and used as an indicator of how much the subjects were able to
limit their movement based on this feedback. In addition, the mean frequency error (i.e. the
difference between the subjects’ actual frequency of stepping and 0.5Hz) was calculated for
each trial as well as the mean across steps of the feet ML distances during the double-stance
phases. This mean ML feet distance and the mean frequency error were used as an indicator of
the accuracy of the subjects’ in performing tandem-walking. The parameters were averaged
across the two cross-over groups to minimize the influence of possible order effect.
Linear regression was used to determine the statistical significance of change in the
across practice trials. Simple paired t tests were also used to determine any significant shortterm retention effect in terms of percentage change of the parameters between before
and after each practice sessions (with and without BF). The parameters were considered
independent for statistical purposes since they were obtained from independent measures,
as a consequence Bonferroni correction was used only when paired t test were repeatedly
applied to the same set of parameters. T-tests verified also that the percentage change of
each parameter occurred across the two sessions of the experiment was not statistically
significantly different between the two groups. In other words, that the changes in the
parameters occurred after the subjects were exposed to both the session of the experiment
were not significantly different.
151
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Results
Immediate Effects of BF
All subjects in both groups improved their stability as soon as the BF device was turned on.
Specifically, in the first three trials with BF, COM displacement was significantly reduced by3.8%,
trunk tilt by 17.8%, and mean ML feet distance by 20% compared to the previous three trials
without BF. Frequency error was the only parameter that increased (by 34.5%) when BF was
turned on. Figure 3 shows raw data of COM displacement and trunk tilt from one representative
subject with and without BF. Note the decrease in variability and amplitude of COM displacement
and trunk tilt occurring with BF.
B. Learning Occurring During an Experimental Session
100
ML COM [mm]
ML COM [mm]
A. Immediate Effect of BF
Without
0
-100
With
0
4
8
12
16
100
Beginning
End
0
-100
0
8
4
10
Without
0
With
-10
0
4
8
12
16
Time [s]
ML Tilt [degree]
ML Tilt [degree]
Time [s]
12
10
Beginning
0
End
-10
0
16
8
4
12
16
Time [s]
Time [s]
Figure 3 – Panel A shows the immediate effect of BF on lateral trunk tilt and COM displacement from one
representative subject. Panel B shows the effect of learning during one experimental session on lateral trunk tilt and
COM displacement from one representative subject.
Table 2 – Parameter at the very beginning (before the first practice session) and at the very end (after the second
practice session). Each value corresponds to the average of three trials.
COM SD [mm]
152
Tilt SD [degree]
Mean feet distance [mm]
Freq. Error [Hz]
Subject #
Beginning
End
Beginning
End
Beginning
End
Beginning
End
1
44.89
39.65
4.85
3.38
64.71
34.70
0.38
0.01
2
57.01
53.48
5.73
4.25
67.46
53.52
0.13
0.07
3
83.06
35.85
6.92
5.36
71.59
68.90
0.16
0.01
4
51.45
28.31
6.47
2.82
76.14
33.71
0.09
0.01
5
65.67
27.90
4.18
2.85
79.71
48.26
0.33
0.09
6
73.03
62.25
3.84
3.32
59.91
66.98
0.32
0.19
7
27.59
27.23
2.63
2.24
48.45
39.05
0.08
0.02
8
47.66
25.17
2.78
1.75
37.27
36.05
0.08
0.14
9
49.27
41.09
3.25
2.32
70.39
43.47
0.24
0.05
Mean(SD)
55.5(16.5)
37.9(12.9)
4.52(1.58)
3.14(1.11)
64.0(13.6)
47.1(13.4)
0.20(0.12)
0.07(0.07)
Chapter 9
Effect of Practicing Tandem Gait on the 4 Parameters Anlyzed
Without BF
With BF
50
40
B.
4.50
3.50
2.50
30
Trials
Trials
60
Without BF
With BF
C.
50
40
30
Trials
Without BF
With BF
0.16
Frequency Error [Hz]
ML Step Variance [mm]
Without BF
With BF
5.50
A.
ML Tilt SD [degree]
ML COM SD [mm]
60
D.
0.12
0.08
0.04
0.00
Trials
Figure 4 – Effect of practicing tandem gait across trials. Each value represents the average among the subjects of
three consecutive trials. Error bars represent standard errors.
Effects of Practicing Tandem Gait
During the experiment, all subjects improved their stability with repetition of tandem
gait trials; COM displacement, trunk tilt SD, mean ML feet distance, and frequency error were
significantly lower for all subjects in the three trials recorded after the two practice sessions than
in the three trials recorded before the two practice sessions (average values are reported in Table
2). Figure 3B shows some raw data of COM displacement and trunk tilt from one representative
subject in the first and last trial of the first experimental sessions. Note the decrease in variability
and amplitude of trunk tilt and COM displacement occurring with practice.
Effects of Practicing Tandem Gait in the Session without BF
The subjects’ COM displacement SD significantly decreased over the course of the session
without BF (Fig. 4A). The regression slope was negative and differed significantly from zero
(p<0.05), and the linear regression accounted for 70% of the variance. The subjects’ SD of trunk
tilt also showed significant reduction while practicing without BF (Fig. 4B). The slope of the linear
regression coefficient of the tilt SD values across trials was negative and was significantly different
from zero (p<0.05), and the linear regression accounted for 60% of the total variation. Subjects’
mean ML feet distance (Fig. 4C) also significantly decreased over time while practicing tandem
153
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Table 3 – Short-term retention after practicing tandem gait without and with BF. Each value represents the
percentage change occurred between before and after each practice session. * indicate statistical significance of
the overall mean percentage change.
COM SD %
Subject #
Tilt SD %
Without
With
Without
1
-4.40
13.61
2
-18.65
11.14
Mean feet distance %
Freq. Error %
With
Without
With
Without
With
-37.34
-7.98
-44.67
-13.74
-86.94
-78.89
-22.31
67.03
-25.12
1.76
-71.53
29.26
3
2.40
-9.06
-44.51
-25.83
-64.48
14.40
-91.70
-89.00
4
-32.99
-23.08
-61.07
-26.22
-41.76
14.86
-64.40
-81.27
5
-16.40
-39.05
-44.65
8.88
-61.57
-11.40
-98.06
38.00
6
-7.24
-19.86
-31.83
15.83
-23.38
4.07
-49.50
-13.57
7
-15.21
-1.27
-22.85
5.34
-11.57
-6.32
-84.96
-65.50
8
-4.03
-14.20
-23.70
-1.19
-25.86
3.68
87.60
-29.42
9
-2.31
-9.80
-12.02
-31.44
-3.27
8.30
-41.43
-69.82
Mean(SD)
*-10.9(10.9)
-10.2(16.6)
*-33.4(15.0)
0.5(30.1)
*-33.5(21.1)
1.7(10.4)
*-55.7(57.0))
*-40.0(48.5)
gait without BF, with 60% of the variance accounted for by the linear regression. The subjects’
stepping frequency error, however, did not significantly change across trials by practicing tandem
gait without BF (p>0.05; Fig4D).
Effects of Practicing Tandem Gait in the Session with BF
During the session with BF, subjects consistently exhibited a smaller trunk COM displacement
SD, tilt SD, and mean ML feet distance than in the session without BF (p<0.05; Figure 4A-C).
Although the linear regression slope was negative for these three parameters, the statistically
analysis showed that the regression slope was not significantly different from zero. The variance
accounted for by the linear regression was 17% for COM displacement, 30% for trunk tilt, and
6% for mean ML feet distance. In contrast, the step frequency error did improve with practice
(regression slope negative and significantly different from zero, p<0.01). The variance accounted
for by the linear regression was 75% for step frequency error.
Short-term Retention Effect of Practicing Tandem Gait
Short-term retention, i.e. the difference between the performances at the beginning and
at the end of each session, was higher after practicing without BF than after practicing with BF.
Table 3 reports the percentage changes between the averages of the first and last three trials
of each session for each parameter. Practicing without BF the significantly reduced trunk COM
displacement SD, tilt SD, mean ML feet distance, and step frequency error in performing tandem
gait (Table 3). Practicing with BF, only frequency error showed significant improvements (Table
3). COM displacement SD decreased for most of the subjects after practicing with BF, however
this change was not significant (p=0.08; Table 3).
154
Chapter 9
Discussion
Motor Learning During Tandem Gait Practice
After practicing tandem gait, all UVL subjects improved their performance in terms of
postural stability and stepping accuracy. These improvements included 1) an increased ML
stability, shown by the reduction of the trunk tilt and COM SD; and 2) a higher accuracy in
maintaining the tandem position of the feet while walking, shown by the reduction of the ML
variability of stepping, as well as a high accuracy in stepping to the metronome rhythm. These
results suggest that with practice, subjects with UVL can learn to better control their posture
during a complex task such as tandem gait. In fact, the lower variability of lateral stepping
placement represents reduced stepping deviation toward the affected side which is a typical
clinical syndrome of UVL subjects [25] [6]. This improved tandem stepping performance may be
due to reduced vestibular-somatosensory conflict and/or increased gain of the proprioceptive
postural loop [26] or to improved feedforward control of the complex multi-segmental task
[27].
Practice Sessions with and without BF
Thanks to the cross-over design adopted for this experiment, we were able to cancel out
the potential effect of session order by averaging across sessions (with and without BF). In
other words, the results reported in Figure 4 are not influenced by the order effect of trials
with and without BF so that the effect of spontaneous learning, occurring when repeating a
task, was equally divided between the 2 sessions. Most previous studies of the effects of BF on
postural control did not control for such a practice affect and attributed all of the improvement
in performance to effects of BF [11].
During trials with BF, all subjects consistently achieved better performances than in trials
without BF. In particular, trunk stability and stepping accuracy were better in trials with BF than
in trials without BF. These results suggest that UVL subjects were able to effectively use BF to
improve their performance during tandem gait consistent with previous studies with other,
less dynamic tasks such as stance [9;28]. Furthermore, this improved performance occurred at
the start of the very first trials with the BF device and did not require a period of practice to
be effective. This immediate improvement of postural control with BF is consistent with our
previous studies of effects of audio-biofeedback on stance posture in subjects with bilateral
vestibular loss and controls [10;29]. During the practice trials in the session with BF, UVL subjects
155
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
did not increase their relative stability as much as during the practice trials without BF. This result
was probably due to the significantly greater stability level induced by the BF leaving a smaller
potential for additional improvement (a floor effect). However, BF consistently improved the
accuracy of the tandem gait performance across practice trials. Specifically, the frequency error
was initially larger in trials with BF than in trials without BF although, in the end, the error was
significantly lower (Fig. 4D). The higher error in frequency of stepping shown at the beginning
of the session with BF may be due to the subjects’ initial inability to pay enough attention to
the metronome and the BF at the same time. Over time, however, all subjects could decrease
this error to the point that they achieved the best performance, in terms of frequency error, in
the trials with BF. This particular result suggests that the use of BF becomes more automatic (i.e.
requires less attention) with practice [30].
Short-term Retention of Motor Learning
Immediately after practice, subjects retained their performance improvements achieved
by practicing tandem gait without BF in terms of trunk stability and accuracy of foot placement,
as shown by the four parameters analyzed in Table 3. This result is further evidence of the
extensive potential for motor learning in UVL subjects [31;32]. Only limited short-term retention
effects were evident after practice in the session with BF. Only one out of four parameters, the
frequency error, was found to retain significant improvements without BF, after practicing with
BF (Table 3). This result may suggest that, immediately after turning the BF device off, subjects
retained a higher level of cognitive attention; attention that they then focused upon the only
remaining external cue, the metronome beat. As a consequence, they more accurately controlled
the frequency of stepping.
Three factors may have limited short-term retention of performance in the other three
parameters (tilt SD, COM SD, and mean ML feet distance) after practice with BF: 1) the short
duration (about 10 minutes) of the practice; 2) the greater number of trials performed without
BF (30) than with BF (24); in fact, tandem gait without BF was both the task for practicing and
for verifying retention of performance; and 3) the experimental protocol was not purposely
designed to facilitate transfer and retention of postural performance. To be more effective, the
protocol could have alternated trials with BF and without BF so that, at the beginning, trials with
BF were more frequent, and then, over time, trials with BF were gradually diminished [33].
Conclusions
UVL subjects can integrate vibrotactile BF information in their postural control to effectively
improve stability and performance accuracy during tandem gait. This improvement occurs as
soon as the BF device is turned on and does not require a period of practice. However, this
integration of augmented sensory information becomes more automatic with practice over
time. Thus, vibrotactile BF acts similarly to natural sensory feedback in improving dynamic motor
156
Chapter 9
performance and not as a method to recalibrate motor performance to improve function
after short-term use.
157
Effect of Trunk-Tilt Tactile Biofeedback on Tandem Gait in Vestibular Loss Subjects
Bibliography
Bibliography
[1]
F. B. Horak and Macpherson J.M., “Postural equilibrium and orientation,” in Handbook of physiology. Rowell
RB and Shepherd J.T., Eds. New York: Published for the American Physiology Society by Oxford University
Press, 1996, pp. 255-292.
[2]
F. B. Horak and C. L. Shupert, “Role of the vestibular system in postural control,” in Vestibular Rehabilitation.
S. J. Herdman, Ed. Philadelphia: F. A. Davis Company, 1994, pp. 22-46.
[3]
M. Lacour, J. Barthelemy, L. Borel, J. Magnan, C. Xerri, A. Chays, and M. Ouaknine, “Sensory strategies in
human postural control before and after unilateral vestibular neurotomy,” Exp. Brain Res, vol. 115, no. 2,
pp. 300-310, June1997.
[4]
C. Parietti-Winkler, G. C. Gauchard, C. Simon, and P. P. Perrin, “Sensorimotor postural rearrangement after
unilateral vestibular deafferentation in patients with acoustic neuroma,” Neurosci. Res., vol. 55, no. 2, pp.
171-181, June2006.
[5]
C. L. Shupert, F. B. Horak, and F. O. Black, “Hip sway associated with vestibulopathy,” J. Vestib. Res, vol. 4, no.
3, pp. 231-244, May1994.
[6]
A. Shumway-Cook and F. B. Horak, “Assessing the influence of sensory interaction of balance. Suggestion
from the field,” Phys. Ther., vol. 66, no. 10, pp. 1548-1550, Oct.1986.
[7]
M. Schwarz and Olson P.R., “A hystorical perspective on the field of biofeedback and applied
psychophysiology,” in Biofeedback: a practitioner’s guide, 3rd ed. M. Schwarz and F. Andrasik, Eds. New
York: Guilford, 2003, pp. 3-19.
[8]
S. T. Moore and M. Woollacott, “The use of biofeedback to improve postural stability,” Phys. Ther. Practice,
vol. 2, pp. 1-19, 1993.
[9]
C. Wall, III and E. Kentala, “Control of sway using vibrotactile feedback of body tilt in patients with moderate
and severe postural control deficits,” J. Vestib. Res., vol. 15, no. 5-6, pp. 313-325, 2005.
[10] M. Dozza, L. Chiari, B. Chan, L. Rocchi, F. B. Horak, and A. Cappello, “Influence of a portable audio-biofeedback
device on structural properties of postural sway,” J. Neuroengineering. Rehabil., vol. 2, p. 13, May2005.
[11] M. Tyler, Y. Danilov, and Bach-y-Rita P, “Closing an open-loop control system: vestibular substitution through
the tongue,” J. Integr. Neurosci., vol. 2, no. 2, pp. 159-164, Dec.2003.
[12] H. Petersen, M. Magnusson, R. Johansson, and P. A. Fransson, “Auditory feedback regulation of perturbed
stance in stroke patients,” Scand. J. Rehabil. Med., vol. 28, no. 4, pp. 217-223, Dec.1996.
[13] K. F. Hamann and C. Krausen, “Clinical applications of posturography - Body tracking and biofeedback
training,” in Disorders of Posture and Gait. T. Brandt, W. Paulus, and W. Bles, Eds. New York: Georg Thieme
Verlag Stuttgart, 1990, pp. 295-298.
[14] D. A. Winter, C. D. MacKinnon, G. K. Ruder, and C. Wieman, “An integrated EMG/biomechanical model of
upper body balance and posture during human gait,” Prog. Brain Res., vol. 97:359-67., pp. 359-367, 1993.
[15] R. D. Easton, A. J. Greene, P. DiZio, and J. R. Lackner, “Auditory cues for orientation and postural control in
sighted and congenitally blind people,” Exp. Brain Res, vol. 118, no. 4, pp. 541-550, Feb.1998.
[16] G. Wu, “Real-time feedback of body center of gravity for postural training of elderly patients with peripheral
neuropathy,” IEEE Trans. Rehabil. Eng, vol. 5, no. 4, pp. 399-402, Dec.1997.
158
Chapter 9
[17] C. Wall, III, M. S. Weinberg, P. B. Schmidt, and D. E. Krebs, “Balance prosthesis based on micromechanical sensors
using vibrotactile feedback of tilt,” IEEE Trans. Biomed. Eng, vol. 48, no. 10, pp. 1153-1161, Oct.2001.
[18] H. Huang, S. L. Wolf, and J. He, “Recent developments in biofeedback for neuromotor rehabilitation,” J.
Neuroengineering. Rehabil., vol. 3:11., p. 11, June2006.
[19] S. L. Wolf, “Electromyographic biofeedback applications to stroke patients. A critical review,” Phys. Ther.,
vol. 63, no. 9, pp. 1448-1459, Sept.1983.
[20] M. S. Weinberg, C. Wall, J. Robertsson, E. O’neil, K. Sienko, and R. Fields, “Tilt Determination in MEMS Inertial
Vestibular Prosthesis,” J. Biomech. Eng., vol. 128, no. 6, pp. 943-956, Dec.2006.
[21] R. J. Peterka, C. Wall, III, and E. Kentala, “Determining the effectiveness of a vibrotactile balance prosthesis,”
J. Vestib. Res., vol. 16, no. 1-2, pp. 45-56, 2006.
[22] C. L. Vaughan, B. L. Davis, and J. C. O’Connor, Dynamics of human gait Kiboho, 1999.
[23] R. F. Chandler, C. E. Clauser, J. T. McConville, J. T. Reynolds, and H. M. Young, Investigation of inertial properties
of the human body. Washington D.C.: Aerospace Medical Research Laboratory, 1975.
[24] W. T. Dempster, “Space Requirements of the Seated Operator,” in Biomechanics and motor control of human
movement. John Wiley & Sons, Ed. 1955.
[25] I. S. Curthoys and G. M. Halmagyi, “Vestibular compensation: a review of the oculomotor, neural, and clinical
consequences of unilateral vestibular loss,” J. Vestib. Res. 1995. Mar. -Apr, pp. 67-107, 1995.
[26] A. M. Bronstein, J. D. Hood, M. A. Gresty, and C. Panagi, “Visual control of balance in cerebellar and
parkinsonian syndromes,” Brain, vol. 113 ( Pt 3), pp. 767-779, June1990.
[27] S. Corna, A. Nardone, A. Prestinari, M. Galante, M. Grasso, and M. Schieppati, “Comparison of CawthorneCooksey exercises and sinusoidal support surface translations to improve balance in patients with unilateral
vestibular deficit,” Arch. Phys. Med. Rehabil., vol. 84, no. 8, pp. 1173-1184, Aug.2003.
[28] E. Kentala, J. Vivas, and C. Wall, III, “Reduction of postural sway by use of a vibrotactile balance prosthesis
prototype in subjects with vestibular deficits,” Ann. Otol. Rhinol. Laryngol., vol. 112, no. 5, pp. 404-409,
May2003.
[29] M. Dozza, F. Horak, and Chiari L., “Auditory Biofeedback Substitutes for Loss of Sensory Information in
Maintaining Stance,” Exp. Brain Res, 2006.
[30] E. De Visser, R. P. Veth, H. W. Schreuder, J. Duysens, and T. Mulder, “Reorganization of gait after limb-saving
surgery of the lower limb,” Am. J. Phys. Med. Rehabil., vol. 82, no. 11, pp. 825-831, Nov.2003.
[31] H. Cohen, “Vestibular rehabilitation improves daily life function,” Am. J. Occup. Ther., vol. 48, no. 10, pp.
919-925, Oct.1994.
[32] A. Shumway-Cook and F. B. Horak, “Rehabilitation strategies for patients with vestibular deficits,” Neurol.
Clin., vol. 8, no. 2, pp. 441-457, May1990.
[33] R. A. Schmidt, Motor control and leaning: a behaviral emphasis. Champaign IL: Human Kinetic Pub, 1982.
159
Conclusions
160
Chapter
Chapter 10
Conclusions: Effects of
Biofeedback on Postural Control
and Potential Impact in Motor
Rehabilitation Therapy
161
Conclusions
162
Chapter 10
Conclusions
Sensory integration is fundamental for the control of posture (see Chapter 1). When
sensory information is reduced, such as in vestibular loss subjects, sensory integration
is impaired. One way to increase sensory information for the control of posture is using
biofeedback devices. The design of biofeedback devices, as well as the design of the
experimental protocols aimed at evaluating the effect of biofeedback on learning and
retention of postural control performances, face many challenges. Specifically, a two doubleblinded, randomized experimental design with both dynamic tasks and static tasks, seems to
be the best protocol to determine the effectiveness and potential impact in the rehabilitation
field of biofeedback systems.
Visual-biofeedback of center-of-pressure displacement is the biofeedback system that
traditionally has received the most interest for experimentation on postural control, and it is
currently used for balance rehabilitation in stance. During stance, trunk acceleration is highly
correlated with center-of-pressure displacement (see Chapter 2). Thus, an audio-biofeedback
system, coding trunk acceleration into a stereo sound modulation, may be an alternative to
visual-biofeedback of center-of-pressure displacement. This new, audio-biofeedback device
is lighter and more cost-effective than traditional visual-biofeedback systems; further it
is portable, so it can be used also during complex dynamic tasks, and does not take over
vision.
Using this audio-biofeedback, healthy subjects reduced sway by increasing control of
posture when sensory information available is limited (see Chapter 3). More specifically, this
sway reduction occurred without increasing muscle activity or muscle co-contraction and was
caused by an enhancement of the closed-loop control of posture. Furthermore, the effect of
audio-biofeedback on postural sway was found to be direction-specific (see Chapter 4).
Also, bilateral vestibular loss subjects reduced sway by increasing control of posture
when sensory information was limited using this audio-biofeedback (see Chapter 5). Further
more, bilateral vestibular loss subjects could take advantage of audio-biofeedback more than
controls when visual and somatosensory information were limited. In addition, the benefit that
each subject could take from audio-biofeedback, was related to their relative dependence on
visual, somatosensory, and vestibular information (see Chapter 6) suggesting use of audiobiofeedback specifically compensates for lack of vestibular, somatosensory, and visual sensory
information.
163
Conclusions
The efficacy of biofeedback and the strategy of postural responses evoked by the
biofeedback were found to depend on the biofeedback design (see Chapter 7). In fact,
depending on the representation and coding of the feedback variable, users were able to
achieve a different performance level and favor a different postural strategy in response to
the biofeedback.
Audio-biofeedback of center-of-pressure improved balance also during dynamic tasks
such as stance perturbed by continuously, randomly oscillating surface (see Chapter 8).
The amount of information from biofeedback needed by a subject to improve balance was
found to depend on the challenge of the task. With practice of stance on a moving surface,
motor learning improved subjects’ postural responses. However, even after practicing, audiobiofeedback continued to be effective in reducing postural responses at low frequencies
(<0.8Hz), suggesting that, with simple motor learning, subjects are not capable to achieve
the same level of performance as with audio-biofeedback.
Another dynamic task, tandem gait, was used to determine the effects of a tactilebiofeedback in subjects with unilateral vestibular loss. Tactile-biofeedback on the lateral
trunk to indicate lateral postural sway was found to improve subjects’ performance while
practicing tandem gait (see Chapter 9). However, one session of practice with biofeedback
did not result in many after-effects consistent with retention of motor performance without
this additional biofeedback. Our results suggest that tactile-biofeedback in tandem gait acts
similar to natural sensory feedback in immediately improving dynamic motor performance
and not as a method to recalibrate motor performance to improve dynamic balance function
after short-term use.
The results and conclusions reported above constitute a brief summary of this thesis.
The above-mentioned results show how different biofeedback designs were found to improve
balance and motor performance in different postural static and dynamic tasks. During this
experimentation, we showed how crucial is the design of a biofeedback system since it
determines 1) the improvement that subjects will be able to achieve and 2) which postural
strategy will be responsible for this improvement. As a consequence, in order to achieve the
best postural performance without eliciting erroneous strategies for the control of posture,
the biofeedback design should be customized for each subject and task. Further, results from
practicing with biofeedback suggest that biofeedback 1) can still be useful after spontaneous
learning occurs and 2) may favor motor learning. However, a customized protocol is necessary
to maximize balance improvement and its potential retention for rehabilitation. Finally, the
findings presented in this thesis constitute clear evidence that biofeedback 1) can increase
basic knowledge about sensory integration and motor control, 2) has the potential, once
conveniently customized, to became a helpful tool for balance and motor rehabilitation and
training, and 3) needs to be equipped with training protocols able to favor motor learning
and control for erroneous control of posture.
164
Chapter 10
The conclusions presented in this thesis, foresee two promising areas of interest for
further study on biofeedback, one related to the biofeedback design and one related to
the biofeedback application. In particular, biofeedback design can be improved by using
virtual reality. Indeed, virtual reality is in essence immersive, multi-modal, attractive, easy-tounderstand, intuitive, and entertaining; further, it permits to recreate real life situations. Such
features are highly desirable in a biofeedback system for augmenting subjects’ motivation
and attention which are known to favor brain plasticity and for testing biofeedback in real,
controlled daily-life situation. Another promising area of interest for studies on biofeedback
regards its experimentation on other classes of subjects with motor impairments, such as
Parkinson’s or after-stroke subjects. In fact, in this context, biofeedback experimentation
could both help understanding the extent to which motor impairments are related to
sensory integration deficits (which, to date, is not totally understood for these classes of
subject) and help increasing the quality of life of these subjects by improving their postural
performances.
165
Conclusions
166
Chapter
Acknowledgements
I would like to thank Dr. Angelo Cappello, Dr. Fay B. Horak, and Dr. Lorenzo Chiari for the
supervision and guide of my PhD program. I also would like to thank Dr. Conrad Wall III, Dr.
Robert J. Peterka, Dr. James S. Frank, and Dr. Frantisek Hlavacka for stimulating inspiration,
assistance, and support in writing some of the papers documenting my PhD study.
Finally, I would also like to thank Dr. Sandra J. Oster, for English editing and education, Dr.
Charles Russell, Andrew Owings, Edward King, and Andrea Sabbioni for technical support, and
Triana Nagel-Nelson, Martina Mancini, Tara Phillips, for assistance during the experiments. This
study was supported by the University of Bologna grant for the PhD program in Bioengineering
and by grants from the NIH DC01849, DC04082, and DC06201.
167

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project