Workshop on Grand Challenge Competition t P di tI Vi KL dt

Workshop on Grand Challenge Competition t P di tI Vi KL dt
Workshop on Grand Challenge Competition
t Predict
to
P di t IIn Vi
Vivo K
Knee L
Loads
d
B.J. Fregly
g y1, Darryl
y D. D’Lima2, and Thor Besier3
1University
of Florida, Gainesville, FL
2Shiley Center at Scripps Clinic, La Jolla, CA
3Stanford University, Stanford, CA
at Scripps Clinic
The Ultimate Goal
•
•
Why are we here this morning?
What do we hope to achieve?
Our ultimate goal is clinical utility of
musculoskeletal computer models.
at Scripps Clinic
Motivation
The Ultimate Goal
Osteoarthritis
Muscle loads
Cartilage loads
Ligament loads
Bone loads
Motion patterns
Stroke
Cerebral palsy
P
Paraplegia
l i
at Scripps Clinic
Motivation
Standard Treatment Design
Currently, treatment design for neuromusculoskeletal
disorders involves the following steps:
1.
2.
3
3.
4.
Observe what has worked well for previous patients.
Create implicit, mental model of patient.
G
Guess
best
b t treatment
t t
t parameters
t
for
f currentt patient.
ti t
Apply treatment and iterate if possible/necessary.
T t
Treatment
t planning
l
i iis hi
highly
hl subjective
bj ti
and outcome is often variable for different patients.
at Scripps Clinic
Motivation
Standard Treatment Design
Currently, treatment design for neuromusculoskeletal
disorders involves the following steps:
1.
2.
3
3.
4.
Observe what has worked well for previous patients.
Create implicit, mental model of patient.
“One size fits none”
G
Guess
best
b t treatment
t t
t parameters
t
for
f currentt patient.
ti t
Apply treatment and iterate if possible/necessary.
T t
Treatment
t planning
l
i iis hi
highly
hl subjective
bj ti
and outcome is often variable for different patients.
at Scripps Clinic
Motivation
Personalized Treatment Design
In the future, treatment design for neuromusculoskeletal
disorders could involve the following steps:
1.
2.
3
3.
4.
Observe what has worked well for previous patients.
Create explicit, computational model of patient.
P f
Perform
virtual
i t l treatments
t t
t on patientpatient
ti t-specific
ifi model.
d l
Apply optimized treatment to patient.
T t
Treatment
t planning
l
i becomes
b
objective
bj ti
and outcome can be optimized for each patient.
at Scripps Clinic
Motivation
Personalized Treatment Design
In the future, treatment design for neuromusculoskeletal
disorders could involve the following steps:
1. Observe what has worked well for previous patients.
The National Academy of Engineering has
2. Create explicit, computational model of patient.
identified “personalized
p
medicine” as one of
3 Perform
3.
P f
virtual
i t l treatments
t t
t on patientpatient
ti t-specific
ifi model.
d l
st
the 10 grand challenges of the 21 century.
4. Apply optimized treatment to patient.
T t
Treatment
t planning
l
i becomes
b
objective
bj ti
and outcome can be optimized for each patient.
at Scripps Clinic
Motivation
Virtual Prototyping
at Scripps Clinic
Motivation
Barriers to Clinical Utility
1)) Model Creation
● Automated patientpatient-specific calibration
● No special engineering/programming skills
● Computationally “fast”
fast
2) Model Utilization
● “Clinically useful locomotion measures”
● Identification of such measures
● Calculation of such measures
3) Model Validation
● Accuracy of calculated measures
● Challenge of unmeasurable quantities
● Limitations in modeling
g capabilities
p
at Scripps Clinic
Motivation
“The
The Emperor’s
Emperor s New Clothes”
Clothes
Do we have a similar phenomenon in the
musculoskeletal modeling community?
• Many publications that predict muscle and
contact forces using unvalidated methods.
• Significant
g
research funding
g going
g g to
projects that are making unvalidated
predictions.
• Statements being made about clinical
conditions and treatments based on
unvalidated predictions.
at Scripps Clinic
Motivation
“The
The Emperor’s
Emperor s New Clothes”
Clothes
Do we have a similar phenomenon in the
musculoskeletal modeling community?
• Many publications that predict muscle and
Of course, the contact
answer
depends
in part onmethods.
forces
using unvalidated
the q
question we
y research
g to answer,
• Significant
gare trying
funding
g going
gbutg to
projects
that areofmaking
unvalidated
should we be more
critical
our own
work?
predictions.
• Statements being made about clinical
conditions and treatments based on
unvalidated predictions.
at Scripps Clinic
Motivation
Workshop Objective
To introduce you to a “grand challenge” competition
competition, to
be held next summer at the SBC, to critically evaluate
in vivo muscle and contact force predictions at the knee
during gait using data collected from a patient with a
force--measuring knee replacement.
force
at Scripps Clinic
Motivation
Big Picture
• We provide the in vivo data (minus the implant loads).
• You predict the muscle and contact forces.
• We evaluate the contact force predictions quantitatively.
• Best predictions are presented in a special session.
• Actual contact forces are revealed in the session.
• Winner is closest to the measured contact forces
forces.
at Scripps Clinic
Motivation
Rationale
In vivo measurement of muscle forces would be
required for direct quantitative validation of muscle
force predictions.
Th
Though
h indirect,
i di t iin vivo
i measurementt off contact
t t fforces
is the next best option for quantitative validation, since
muscle forces are the p
primary
y determinants of jjoint
contact forces (Herzog et al.,
al., 2003).
at Scripps Clinic
Motivation
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
3. Experimental Data Collection (Thor Besier
Besier))
4. Modeling Results To Date (B.J. Fregly
Fregly))
5. Logistics of Competition (Darryl
(Darryl D’Lima)
D’Lima)
6. Questions and Answers (All)
at Scripps Clinic
Reminder
Please sign the attendance sheet if you
want to receive ee-mail updates about
organization of the competition.
at Scripps Clinic
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
at Scripps Clinic
2. Instrumented Implant
D i and
Design
dA
Accuracy
Darryl
y D. D’Lima,
D’Lima, M.D., Ph.D.
Director, Orthopaedic Research Laboratories
Shiley Center for Orthopaedic Research & Education
pp Clinic,, La Jolla,, CA
Scripps
at Scripps Clinic
Generation I Tray Design
•
Axial Load Cells (4)
– Total Load
– Mediolateral Distribution
– Center of Pressure
– AP/ML Moments
– Shear
– Axial Moment
at Scripps Clinic
2. Implant Design and Accuracy
Generation I Tray Design
•
“eKnee”
Axial Load Cells (4)
– Total Load
– Mediolateral Distribution
– Center of Pressure
– AP/ML Moments
– Shear
– Axial Moment
at Scripps Clinic
2. Implant Design and Accuracy
Generation I Tray Design
at Scripps Clinic
2. Implant Design and Accuracy
Generation I
Calibration Accuracy
•
•
•
•
•
NIST Load cell
R2 > 0.99
AAE Axial Force < 1.1% FS
Shear cross-talk
cross talk <0.3%
<0 3%
AAE Center of Pressure <0.25 mm
Kaufman +,
+ J Biomech 1996
at Scripps Clinic
2. Implant Design and Accuracy
Generation I
Calibration Accuracy
at Scripps Clinic
2. Implant Design and Accuracy
Generation I
Calibration Accuracy
•
•
•
•
NIST Load cell
R2 > 0.99
AAE Axial Force < 1.5% FS
AAE Center of Pressure < 1.9 mm
D’Lima +, J Biomech 2005
at Scripps Clinic
2. Implant Design and Accuracy
Generation II Tray Design
Microprocessor
Internal
Power Induction
Coilil
I t
lP
I d ti C
Transmitting
g Antenna
Kirking +, J Biomech, 2005
at Scripps Clinic
2. Implant Design and Accuracy
Generation II Tray Design
Microprocessor
“eTibia”
Internal
Power Induction
Coilil
I t
lP
I d ti C
Transmitting
g Antenna
Kirking +, J Biomech, 2005
at Scripps Clinic
2. Implant Design and Accuracy
Generation II
Calibration Accuracy
Kirking +, J Biomech 2006
at Scripps Clinic
2. Implant Design and Accuracy
Generation II
Calibration Accuracy
Kirking +, J Biomech 2006
at Scripps Clinic
2. Implant Design and Accuracy
Generation II
Calibration Accuracy
Kirking +, J Biomech 2006
at Scripps Clinic
2. Implant Design and Accuracy
Temperature Tests
•
•
at Scripps Clinic
Water Bath 42
42°°C
High Temperature Burn
Burn--In 80
80°°C
2. Implant Design and Accuracy
Durability Tests
•
•
at Scripps Clinic
Shaker Tests
Prototypes & Implantable Grade Units
– +12 years
2. Implant Design and Accuracy
Data Transmission
•
•
•
•
•
•
at Scripps Clinic
Power Channel
Temperature Channel
12 Data Channels
St
Startt b
byte
t
Checksum byte
2 ms delay
2. Implant Design and Accuracy
Conclusions
1. High sensor accuracy
2. Robust measurements
3 Consistent in vivo measurements
3.
at Scripps Clinic
2. Implant Design and Accuracy
Acknowledgments
SCORE
Clifford Colwell, MD
Shantanu Patil, MD
Juan Hermida, MD
Nick Steklov
Microstrain
Steve Arms
Christopher Townsend
D’Lima
OREF 2609
NIH R21 EB004581
NIH R21 AR057561
SCORE
Zimmer, Inc
Janet Krevolin
Todd Johnson
at Scripps Clinic
2. Implant Design and Accuracy
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
3. Experimental Data Collection (Thor Besier
Besier))
at Scripps Clinic
3. Experimental Data Collection
Thor Besier,
Besier, Ph.D.
Research Director, Human Performance Lab
Department of Orthopaedics
Stanford University,
y, Stanford,, CA
at Scripps Clinic
Organizers
Main Organizers
• Darryl D’Lima,
D’Lima, Shiley Center at Scripps Clinic
• B.J
B.J.. Fregly,
Fregly, University of Florida
EMG Data
• Thor
Th Besier
B
Besier,
i , Stanford
St f d University
U i
it
• David Lloyd, University of Western Australia
Strength Data
• Marcus Pandy,
Pandy, University of Melbourne
at Scripps Clinic
3. Experimental Data Collection
Subject Description
•
•
•
•
•
•
•
at Scripps Clinic
Subject: JW
Gender: Male
A
Age:
83 yrs
Height: 166 cm
Mass: 64
64.6
6 kg
Right knee, generation I implant design
Anthropometric
p
data available from
calibrated subjectsubject-specific skeletal
model ((Reinbolt
Reinbolt et al.,
al., 2008)
3. Experimental Data Collection
Session Description
• Gait
G it and
d other
th motion
ti d
data
t collected
ll t d iin th
the morning.
i
• Strength data collected in the afternoon.
• Fluoroscopic motion data reported previously (Zhao
et al.,
al., 2007).
at Scripps Clinic
3. Experimental Data Collection
Task Summary
•
•
Session 1: Gait
Laboratory
Session 2:
Dynamometer
Laboratory
at Scripps Clinic
St
Static
ti ttrials
i l
Inverse dynamic model calibration
–
Hip, knee, and ankle isolated motion
•
•
•
Musculoskeletal model calibration
Medial
Medial--lateral load manipulation
Gait trials (4 types)
•
Isometric isokinetic,
Isometric,
isokinetic, and passive
dynamometry
3. Experimental Data Collection
Gait Lab Data
•
•
M k ttrajectories
Marker
j t i
– 8-camera Motion Analysis system
– Modified Cleveland Clinic marker set
Ground reaction forces
–
•
•
3 Bertec force plates
Surface
S f
EMG
G
– 14 muscles
– Delsys
y Bagnoli
g
EMG system
y
Joint contact forces
–
at Scripps Clinic
eKnee:: as described previously
eKnee
3. Experimental Data Collection
Dynamometer Lab Data
•
•
Knee flexion angle
– Goniometer & Biodex angle
Joint torque (gravity corrected)
–
•
Surface EMG
–
–
•
Biodex
14 muscles
Delsys Bagnoli EMG system
Joint contact forces
–
at Scripps Clinic
as described previously
Biodex dynamometer
3. Experimental Data Collection
Surface Marker Data
1-2 : Shoulder
3-4 : Elbow
5-6 : Wrist
7-8 : ASIS
9 : Sacrum
10
10--15 : Thigh superior, inferior, lateral
16
16--19: Knee medial and lateral (static only)
20
20--21 : Patella
22
22--27 : Shank superior, inferior, lateral
28
28--31: Ankle medial and lateral (static only)
32
32--33 : Heel
34
34--37 : Midfoot lateral and superior
38
38--39 : Toe tip
40
40--43 : Toes medial and lateral (static
(
only)
y)
at Scripps Clinic
3. Experimental Data Collection
Surface EMG Data
1
1.
2.
3.
4.
5.
6
6.
7.
8.
S i
Semimembranosus
b
Biceps femoris
Vastus medialis
Vastus lateralis
Rectus femoris
femoris**
Medial gastrocnemius
Lateral gastrocnemius
Tensor fascia latae
latae**
9
9.
10.
11.
12.
13.
14
14.
Tibi li anterior
Tibialis
t i
Peroneus longus
Soleus
Adductor magnus
Gluteus maximus
Gluteus medius
medius**
Electrode placement consistent
with Perotto & Delagi (1980)
* Indicates double
double--differential electrode
at Scripps Clinic
3. Experimental Data Collection
EMG Preparation Trials
•
•
Skin shaved and abrased with gauze and then rubbed
with alcohol prior to electrode placement
Manual restraint of subject during maximum isometric
voluntary
l t
contractions
t ti
(3 repetitions):
titi
)
– Hip flexionflexion-extension (standing)
– Knee flexionflexion-extension (seated w knee @ 80
80°°)
– Ankle dorsiflexion (seated w knee @ 40
40°°; ankle @ 0°
0°
–
dorsiflexion)
dorsiflexion)
Ankle plantarflexion (seated w knee @ 40
40°° and standing tiptiptoes)
Ankle inversioninversion-eversion (seated w knee @ 40
40°°)
–
• Resting signals obtained during quiet sitting
at Scripps Clinic
3. Experimental Data Collection
Static Trials
•
•
•
at Scripps Clinic
Standing (toes forward, toes in, toes out)
Sitting
Maximum isometric contraction
3. Experimental Data Collection
Model Calibration Trials
•
•
•
•
•
•
•
at Scripps Clinic
Passive seated leg rest
Unloaded seated leg extension
Loaded seated leg extension
One
One--legged standing
Two--legged squat
Two
Chair rise
Calf raise
3. Experimental Data Collection
Load Manipulation Trials
•
•
at Scripps Clinic
Varus
Varus--valgus stress test
Stance initiation tests
3. Experimental Data Collection
Gait Trials
•
•
•
•
at Scripps Clinic
Normal gait
Medial thrust gait
Walking
gp
pole gait
g
Trunk sway gait
3. Experimental Data Collection
Dynamometer Trials
•
•
at Scripps Clinic
Isometric, passive, and isokinetic knee
flexion/extension
Isometric,, passive,
p
, and isokinetic
ankle plantarflexion
plantarflexion//dorsiflexion
3. Experimental Data Collection
Data To Be Made Available
•
•
•
•
•
EMG preparation trials
Static trials
Model calibration trials
Gait trials
Dynamometer trials
minus the eKnee contact forces for competition
ti l
trials.
at Scripps Clinic
3. Experimental Data Collection
Additional Available Data
•
Pre
Pre-- and postpost-surgery CT scans of knee region
•
Fluoroscopic motion measurements for
treadmill gait (Zhao et al
al.,
., 2007)
at Scripps Clinic
3. Experimental Data Collection
Data Synchronization
Ground Reaction Forces
[3840H ]
[3840Hz]
Joint Contact Forces
[~50Hz]
LP: Low pass cutoff frequency
HP: High pass cutoff frequency
LP:15Hz
Joint Contact Forces
[200Hz]
Common sync signals – vertical GRF and
vastus lateralis EMG
MATLAB
-Cubic spline interpolation
-Cross-correlation
C
l i
-Filtering [4th order Butterworth]
LP:100Hz
HP 30H
HP:30Hz
Ground Reaction Forces
[[1000Hz]]
at Scripps Clinic
EMG
[1000H ]
[1000Hz]
EMG
[[1000Hz]]
3. Experimental Data Collection
Marker Trajectories
[120Hz]
LP:15Hz
Marker Trajectories
j
[200Hz]
Acknowledgments
D’Lima
Fregly
Besier
at Scripps Clinic
3. Experimental Data Collection
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
3. Experimental Data Collection (Thor Besier
Besier))
4. Modeling Results To Date (B.J. Fregly
Fregly))
at Scripps Clinic
4. Modeling Results to Date
B.J. Fregly
Fregly,
g y, Ph.D.
Department of Mechanical & Aerospace Engineering,
Department of Biomedical Engineering, and
Department of Orthopaedics & Rehabilitation
University of Florida, Gainesville, FL
at Scripps Clinic
Previous Studies
1) First eKnee Data Set
Studyy 1 - Correlation between the knee adduction
moment and medial contact force within the gait cycle
Study 2 - Estimation of muscle and contact forces in
the knee during gait
2) Second eKnee Data Set
Study 3 - Do changes in peak knee adduction moment
predict changes in peak medial contact force?
at Scripps Clinic
4. Modeling Results to Date
First eKnee Data Set
• Fluoroscopic
p motion data for treadmill g
gait,, step
p
•
up/down, kneel, and lunge
Video motion and ground reaction data for step
up/down and 5 gait patterns (normal
(normal, fast
fast, slow
slow,
toe out, wide)
at Scripps Clinic
4. Modeling Results to Date
Study 1 Overview
Gait
Analysis
Dynamic
Contact Model
In vivo kinematic
measurement
R
Regression
i
at Scripps Clinic
In vivo load
measurement
Model
Adduction
moment
4. Modeling Results to Date
Medial contact
force
Dynamic Contact Simulation
In vivo knee force data
Dynamic contact model
In vivo knee motion data
at Scripps Clinic
4. Modeling Results to Date
Contact conditions
Dynamic Contact Simulation
Simulation closely matches eKnee total contact
force, eKnee A/P and M/L center of pressure,
and fluoroscopic motion measurements.
In vivo knee force data
Dynamic contact model
In vivo knee motion data
at Scripps Clinic
Contact conditions
Zhao et al., 2007a, Journal of
Orthopaedic Research
4. Modeling Results to Date
Knee Adduction Moment
Are knee adduction moment changes within the
gait cycle highly correlated with changes in
medial contact force?
at Scripps Clinic
4. Modeling Results to Date
Adduction Torque
e (%BW*H)
External
External--Internal Correlation
4
Best
3
Worst
2
1
0
-1
Force (BW)
3
Total
Medial
2
1
0
Medial Force Rattio (%)
75
60
45
30
15
0
0
at Scripps Clinic
20
40
60
Gait Cycle (%)
80
100
0
20
40
60
Gait Cycle (%)
4. Modeling Results to Date
80
100
Correlation Coefficients
Best
Worst
Medial Force (BW)
2
1.5
R = 0.96
p < 0.001
R = 0.83
p < 0.001
1
0.5
0
-1
1
at Scripps Clinic
0
1
2
3
Adduction Torque (%BW*H)
4
-1
1
0
1
2
3
Adduction Torque (%BW*H)
4. Modeling Results to Date
4
Correlation Coefficients
Best
Worst
Medial Fo
orce Ratio (%
%)
75
60
R=0
0.95
95
p < 0.001
R=0
0.73
73
p < 0.001
45
30
15
0
-1
1
0
1
2
3
Adduction Torque (%BW*H)
4
1
-1
0
1
2
3
Adduction Torque (%BW*H)
Zhao et al., 2007b, Journal of
Orthopaedic Research
at Scripps Clinic
4. Modeling Results to Date
4
Correlation Coefficients
Best
Worst
Medial Fo
orce Ratio (%
%)
75
60
R=0
0.95
95
p < 0.001
R=0
0.73
73
p < 0.001
For all 15 trials analyzed together, R = 0.88 for
30
medial force and 0
0.83
83 for medial force ratio
ratio.
45
15
0
-1
1
0
1
2
3
Adduction Torque (%BW*H)
4
1
-1
0
1
2
3
Adduction Torque (%BW*H)
Zhao et al., 2007b, Journal of
Orthopaedic Research
at Scripps Clinic
4. Modeling Results to Date
4
Contact Force Sensitivity
Should highly accurate fluoroscopic kinematic
measurements be directly input into contact
models to calculate in vivo contact forces?
at Scripps Clinic
4. Modeling Results to Date
Contact Force Sensitivity
at Scripps Clinic
4. Modeling Results to Date
Contact Force Sensitivity
Fregly et al., 2008, Journal of
Orthopaedic Research
at Scripps Clinic
4. Modeling Results to Date
Study 2 Overview
Muscle Forces
Geometric Model
Contact Forces
Combined Model
Inverse Dynamic Model
at Scripps Clinic
4. Modeling Results to Date
Muscle & Contact Force Estimation
No contact
at Scripps Clinic
4. Modeling Results to Date
Muscle & Contact Force Estimation
No contact
Contact
Assumptions required
about contact contributions
t inverse
to
i
dynamic
d
i lloads
d
at Scripps Clinic
4. Modeling Results to Date
at Scripps Clinic
Medial Force (N)
M
2500
2000
1500
1000
500
0
Late
eral Force (N)
2500
2000
1500
1000
500
0
Total Force (N)
Sequential Contact Force
2500
2000
1500
1000
500
0
Fast
Normal
Slow
Experiment
Model
0
20 40 60 80 100
Gait Cycle (%)
0
20 40 60 80 100
Gait Cycle (%)
0
20 40 60 80 100
Gait Cycle (%)
4. Modeling Results to Date
Medial Force (N)
M
Sequential Contact Force
Fast
2500
2000
1500
1000
500
0
Normal
Slow
Late
eral Force (N)
2500
2000
1500
1000
500
0
Total Force (N)
Excellent contact force estimates, BUT
lateral collateral ligament tension tuned
to match measured lateral contact forces.
2500
2000
1500
1000
500
0
Experiment
Model
0
20 40 60 80 100
Gait Cycle (%)
0
20 40 60 80 100
Gait Cycle (%)
0
20 40 60 80 100
Gait Cycle (%)
Kim et al.,, 2009,, Journal of
Orthopaedic Research
at Scripps Clinic
4. Modeling Results to Date
Muscle & Contact Force Estimation
Contact
No assumptions required
about contact contributions
t inverse
to
i
dynamic
d
i lloads
d
at Scripps Clinic
4. Modeling Results to Date
Knee Contact Model
+ surrogate contact models of TF and PF joints
at Scripps Clinic
4. Modeling Results to Date
Inverse Dynamic Model
•
•
•
•
•
Full--body model
Full
Three--dimensional
Three
Engineering
g
g jjoints
Calibrated lower body joints
Calibrated full body masses
Reinbolt et al., 2005, Journal of
Biomechanics; Reinbolt et al., 2008,
Medical Engineering & Physics
at Scripps Clinic
4. Modeling Results to Date
Model Registration
at Scripps Clinic
4. Modeling Results to Date
Complete Knee Model
at Scripps Clinic
•
11 muscles controlled by 8
activation signals
•
Muscle force = peak isometric
force x activation
•
Patellar ligament modeled as
3 parallel springs
•
•
Grounded femur
•
6 DOF tibiofemoral joint (3 free
and 3 prescribed DOFs)
6 DOF p
patellofemoral jjoint ((6
free DOFs)
4. Modeling Results to Date
Optimization Problems
“Constrained” formulations – in vivo contact
forces used as additional constraints.
“Unconstrained” formulations – in vivo contact
forces not used as additional constraints.
at Scripps Clinic
4. Modeling Results to Date
Predicted Motion
at Scripps Clinic
4. Modeling Results to Date
Load Decomposition
How do muscle and contact forces contribute
to the six inverse dynamic loads at the knee
during gait?
at Scripps Clinic
4. Modeling Results to Date
300
30
150
15
Tx (Nm)
Fx (N)
Load Decomposition
0
-150
-15
-300
300
-30
30
2000
30
1000
15
Ty
y (Nm)
x
Fy (N)
F
y
0
0
-1000
Net
0
-15
-2000
-30
300
30
150
15
Tz (Nm)
Fz (N)
z
0
-150
-300
at Scripps Clinic
0
25
50
75
Gait cycle (%)
100
0
-15
-30
0
4. Modeling Results to Date
25
50
75
Gait cycle (%)
100
300
30
150
15
Tx (Nm)
Fx (N)
Load Decomposition
0
-150
-15
300
-300
-30
30
2000
30
1000
15
Ty
y (Nm)
x
Fy (N)
F
y
0
0
-1000
Net
Contact
0
-15
-2000
-30
300
30
150
15
Tz (Nm)
Fz (N)
z
0
-150
-300
at Scripps Clinic
0
25
50
75
Gait cycle (%)
100
0
-15
-30
0
4. Modeling Results to Date
25
50
75
Gait cycle (%)
100
300
30
150
15
Tx (Nm)
Fx (N)
Load Decomposition
0
-150
-15
300
-300
-30
30
2000
30
1000
15
Ty
y (Nm)
x
Fy (N)
F
y
0
0
-1000
Net
Contact
Muscle
0
-15
-2000
-30
300
30
150
15
Tz (Nm)
Fz (N)
z
0
-150
-300
0
25
50
75
Gait cycle (%)
100
0
-15
-30
0
25
50
75
Gait cycle (%)
100
Fregly et al., 2009, SBC
at Scripps Clinic
4. Modeling Results to Date
Muscle & Contact Force Estimates
Does inclusion of explicit contact models in
a musculoskeletal knee model improve the
estimation of muscle and contact forces
during gait?
at Scripps Clinic
4. Modeling Results to Date
“Constrained”
Constrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Constrained”
Constrained Contact Forces
100 N
RMSE
at Scripps Clinic
4. Modeling Results to Date
100 N
RMSE
“Constrained”
Constrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Constrained”
Constrained Muscle Forces
at Scripps Clinic
4. Modeling Results to Date
“Constrained”
Constrained Muscle Forces
at Scripps Clinic
4. Modeling Results to Date
“Constrained”
Constrained Muscle Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Contact Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Muscle Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Muscle Forces
at Scripps Clinic
4. Modeling Results to Date
“Unconstrained”
Unconstrained Muscle Forces
Fregly et al., 2009, SBC
at Scripps Clinic
4. Modeling Results to Date
Study 3 Overview
Gait
Analysis
Regression
at Scripps Clinic
In vivo load
measurement
Model
Adduction
Add
ti
moment
Medial
M
di l contact
t t
force
4. Modeling Results to Date
Joint Contact Forces
How do medial thrust and walking pole gait
affect medial and lateral contact force?
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
at Scripps Clinic
4. Modeling Results to Date
Lateral Contact Force
at Scripps Clinic
4. Modeling Results to Date
Lateral Contact Force
at Scripps Clinic
4. Modeling Results to Date
Lateral Contact Force
at Scripps Clinic
4. Modeling Results to Date
Lateral Contact Force
Fregly et al
al., 2009,
2009 Journal of
Orthopaedic Research
at Scripps Clinic
4. Modeling Results to Date
Knee Adduction Moment
Does the knee adduction moment predict no
change in the first peak and a significant reduction
in the second peak of medial contact force?
at Scripps Clinic
4. Modeling Results to Date
Knee Adduction Moment
at Scripps Clinic
4. Modeling Results to Date
Knee Adduction Moment
at Scripps Clinic
4. Modeling Results to Date
Medial Contact Force
Inconsistent with adduction
moment changes !
at Scripps Clinic
4. Modeling Results to Date
Optimal Axial Rotation
Consistent with medial
contact force changes
Optimal rotation:
Optimal R2 value:
at Scripps Clinic
20 deg
0 57
0.57
-5 deg
0 70
0.70
15 deg
0 74
0.74
4. Modeling Results to Date
Knee Extension Moment
Walter et al., 2009, SBC
at Scripps Clinic
4. Modeling Results to Date
Conclusions
1. Inclusion of explicit contact models in a musculo
musculo-skeletal knee model allows additional inverse
dynamic loads to be used as constraints and
alters the muscle and contact force estimates.
2. The second eKnee data set provides the unique
opportunity to evaluate muscle and contact force
predictions for gait patterns that modulate medial
contact force.
at Scripps Clinic
4. Modeling Results to Date
Acknowledgments
NSF CAREER award CBET 0239042 and
NSF award CBET 0602996
at Scripps Clinic
4. Modeling Results to Date
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
3. Experimental Data Collection (Thor Besier
Besier))
4. Modeling Results To Date (B.J. Fregly
Fregly))
5. Logistics of Competition (Darryl
(Darryl D’Lima)
D’Lima)
at Scripps Clinic
5. Logistics of Competition
Darryl
y D. D’Lima,
D’Lima, M.D., Ph.D.
Director, Orthopaedic Research Laboratories
Shiley Center for Orthopaedic Research & Education
pp Clinic,, La Jolla,, CA
Scripps
at Scripps Clinic
Announcement of Competition
Focus on the musculoskeletal modeling community:
• BIOMCH
BIOMCH--L Newsgroup
• ISB Technical Group on Computer Simulation
Newsgroup
• ASME Summer Bioengineering
g
g Conference
• American Society of Biomechanics Newsletter
• International Society of Biomechanics Newsletter
• SimTK.org ee--mail list
• Personal invitation
at Scripps Clinic
5. Logistics of Competition
Journal of Orthopaedic Research
• Publication
– Make data available
– Announce competition
– Peer reviewed
– Tim Wright, PhD (Editor)
• Data
– Anthropometric
A th
t i measurements
t
– Marker positions
– Ground reaction forces
– EMG signals
– Limited tibial contact forces
– OpenSim
p
model of subject
j
and
implant geometry
at Scripps Clinic
5. Logistics of Competition
www.SimTK.org
www SimTK org
•
•
•
•
•
•
at Scripps Clinic
Registration
g
Data published in J Orthop Research
Contact models of implant components
Videos of data collection
Post--competition implant contact forces
Post
Special requests
5. Logistics of Competition
Predicted Quantities
Time histories of
•
•
Medial contact force
Lateral contact force
for selected gait trials
at Scripps Clinic
5. Logistics of Competition
Abstract Submission
•
•
•
•
•
at Scripps Clinic
Introduction
Methods
Results
Discussion
Predictions – upload to SimTK.org
5. Logistics of Competition
Review Criteria
•
•
•
•
•
•
•
at Scripps Clinic
Reviewers
Significance (0(0-3 points)
Technical content (0(0-5 points)
Completeness
p
((0(0-2 p
points))
Accuracy (0(0-5 points - new)
Novelty (0(0-5 points - new)
Max 20 points
5. Logistics of Competition
Special Session
•
•
•
•
•
at Scripps Clinic
Top scoring papers given podium presentations
in a special session at next year’s conference.
More than one special session may be
possible.
Participants present models and predictions.
Actual contact force measurements revealed at
end of special session.
Post--mortem mini
Post
mini--workshop after special
session to evaluate competition and lessons
learned.
5. Logistics of Competition
Award Presentation
•
•
•
•
at Scripps Clinic
Certificate
Cash prize (hopefully)
Manuscript submitted to J Orthop Research
(investigating)
Runners ups
4. Modeling Results to Date
Workshop Outline
1. Motivation for Competition (B.J. Fregly
Fregly))
2 Instrumented Implant Designs and Accuracy
2.
(Darryl D’Lima)
D’Lima)
3. Experimental Data Collection (Thor Besier
Besier))
4. Modeling Results To Date (B.J. Fregly
Fregly))
5. Logistics of Competition (Darryl D’Lima
D’Lima))
6. Questions and Answers (All)
at Scripps Clinic
6. Questions and Answers
B.J. Fregly
Fregly,
g y, Ph.D., Universityy of Florida and
Darryl D’Lima,
D’Lima, M.D., Ph.D., Shiley Center at Scripps Clinic
at Scripps Clinic
Data Related Questions
1. For which tasks should in vivo contact force data be
released BEFORE the competition?
• EMG preparation trials?
• Static trials?
• Model calibration trials?
• Gait trials (4 patterns)?
• Dynamometer
D
t ttrials?
i l ?
at Scripps Clinic
6. Questions and Answers
Data Related Questions
2. Are the current filter cutoff and output frequencies
acceptable for the data?
Experimental
Quantity
Input
Frequency (Hz)
Filter
Frequency (Hz)
Output
Frequency (Hz)
Marker positions
120
Low pass 15
200
eKnee forces
~50
Low pass 15
200
Ground reactions
3840
Low pass 100
1000
EMG signals
1000
High pass 30
1000
at Scripps Clinic
6. Questions and Answers
Model Related Questions
1. Should we provide our surrogate contact model in
Matlab so that every participant can calculate
tibiofemoral and patellofemoral contact forces easily?
2. If so, how should muscle forces be applied to it?
3. Should we provide an
O
OpenSim
Si version
i off the
th
geometric/inverse dynamic
knee model?
4. What other modeling
information is needed?
at Scripps Clinic
6. Questions and Answers
Organization Related Questions
1. Should accuracyy be the p
primary
y scoring
g criterion, or
should the proposed 5 scoring criteria (significance,
technical content, completeness, accuracy, and
novelty) be used?
2. Should selection of the winning paper be subjective
or objective?
j
If subjective,
j
who should do it?
at Scripps Clinic
6. Questions and Answers
Other Relevant Questions
What questions and suggestions do you have for us?
at Scripps Clinic
6. Questions and Answers
Was this manual useful for you? yes no
Thank you for your participation!

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

Download PDF

advertising