PERSONALIZED MODELS AND
HEALTH MAINTENANCE FOR
MOBILITY
Fregly and Rodgers
Overview
Personalized Models
Health Maintenance
Gaps
Summary
Overview
Personalized Models
Motivation
Personalized modeling methods
Full
-
body scan for personalized model creation
Customization of robot function using personalized
models
Health Maintenance
Gaps
Summary
Why Personalized Models?
“One size fits none”
-
everyone is different!
Increases objectivity in treatment planning (different
clinicians may plan different treatments given same
patient data).
Can facilitate identification of previously unknown
treatments (e.g., modified gait to treat knee OA).
May permit identification of best treatment option
for a specific patient.
May permit identification of sensitive treatment
parameters (i.e., which parameter values do
clinicians need to “get right”?)
Highly Variable Outcomes
For some treatments, standard deviation in
outcome is bigger than the effect size.
Where models
should be focused
Where problems are
too complex for models
Scope of model applicability must be
properly defined.
HIGH FIDELITY ANATOMIC
SHOULDER & ELBOW MODEL
Frans C.T. van der Helm, Ph.D., Delft
Delft Shoulder and Elbow Model
High fidelity anatomic musculoskeletal model
constructed from extensive measurements
performed on a single cadaver specimen.
Model accounts for more variables (including
sarcomere
length) than any other upper extremity
model.
Model validation: Muscle forces cannot be
measured, so no strict validation!
The same model personalization approach cannot
be performed on living patients.
Model Applications
Glenohumeral
arthrodesis
Glenohumeral
endoprosthesis
Tendon transfer after brachial plexus lesion
“Reverse” shoulder
endoprosthesis
Scapula fractures
Functional electrical stimulation for
tetraplegics
Neurological disorders
Computer Assisted Surgical Planning (CASP)
Wheelchair propulsion
Garbage collection
Brick
-
layering
ORTHOPAEDIC SURGERY
AND REHABILITATION
Maria Benedetti, M.D., Alberto
Leardini
, Ph.D., and
Marco
Viceconti
, Ph.D.
-
Bologna
Possible Uses of Gait Analysis
Assessment
–
Assess after treatment how the
treatment worked for a group of patients.
(Common)
Identification
–
Identify on an individual patient
basis which patients should be treated (but not how
they should be treated).
(Becoming more common)
Prediction
–
Predict on an individual patient basis
which treatment should be performed and how it
should be performed (where personalized models
may help).
(Does not yet happen)
The potential value of
Prediction
depends on the
clinical problem at hand.
Clinical Example for Prediction
Clinical Situation:
Oncological
patients who receive
a limb salvage procedure.
Problem:
How to get the bone allograft to heal
–
it
needs load to repair but not so much that it breaks.
Observation:
Each case is unique
–
surgical and
rehabilitation design are not stereotypical.
Proposed solution:
Treatment design using gait and
imaging data in a personalized musculoskeletal
model that estimates muscle & bone loads.
Challenge:
How to gain confidence in patient
-
specific predictions of muscle & bone loads?
Personalized Model Creation & Use
Valente et al.,
Computer Aided Medicine Conference
, 2010
Design of Total Ankle Replacement
Though not personalized, design developed using
patient data and modeling methods.
Leardini
et al.,
Clin
Orthop
Relat
Res
, 2004
PATIENT
-
SPECIFIC
MUSCULOSKELETAL MODELS
Bart
Koopman
, Ph.D. and Herman van der
Kooij
,
Ph.D.,
Enschede
Model Personalization
Problem:
Most musculoskeletal models are generic,
and uniform scaling is inaccurate.
Solution:
Scale/deform a generic parametric model
to match each patient.
Image based scaling of bone geometry (CT, MRI)
Functional kinematic scaling of joint positions/orientations
(marker
-
based motion, laser scans, inertial sensors)
Functional dynamic scaling of muscle strength
(dynamometers)
Challenge:
Fusion of data from
different modalities.
Model Utilization
Collect
pre
-
treatment imaging, kinematic, and
dynamic data.
Simulate
surgical scenarios and parameters.
Select
scenario and parameters that optimize post
-
treatment outcome.
Implement
plan in surgical navigation system.
Validate
model predictions using surgical cases not
planned with model.
Example:
Which tendon to transfer
to restore hip abduction strength in
patients with
Trendelenburg
gait?
Neuro
-
Mechanical Models
How do people interact with and adapt to their
environments (e.g., with robotic systems)?
Visual,
proprioceptive
, and vestibular feedback all
play a role.
No simulation tools currently exist to optimize
human
-
machine interactions in rehabilitation devices.
3D PERSONALIZED
MUSCULOSKELETAL MODELS
Waffa
Skalli
, Ph.D., Paris
Research Goal
Multiscale
personalized human musculoskeletal models
that enable:
Early detection of balance abnormalities.
Design of innovative devices for prevention and
treatment of musculoskeletal disorders.
Identification of the source of pathology (e.g., is it
muscular or skeletal?).
Quantitative assessment of
treatment strategies.
Personalized Modeling
Personalized spine models
for studying scoliosis
Partners: Hospitals in Paris,
Saint Etienne, and Montreal
Biplane X
-
ray Modeling Technology
Internal
-
External Registration
Direct
registration of
presonalized
skeletal
models
to
external
marker locations for
gait
analysis
.
Bi
-
plane X
-
rays with External Markers
Gait Data
Where are the bones with respect to the skin markers?
HOCOMA
-
ADVANCED
FUNCTIONAL MOVEMENT
THERAPY
Peter Hostettler, PhD & CEO, and team, Zurich
Future Directions
Neurorehabilitation
is current focus.
Orthopaedic rehabilitation viewed as a potentially
big future market.
Current robotic training system designed using the
gait pattern of one of the designers.
Customization of robot to individual
patients could be valuable in the the
future (with possible role of personalized
personalized
modeling).
Overview
Personalized Models
Health Maintenance
Remote monitoring
Remote training & treatment
Prediction modeling
Gaps
Summary
REMOTE MONITORING AND
REMOTELY SUPERVISED
TRAINING & TREATMENT
Hermie
Hermens
, PhD,
Enschede
Remote Health Care Vision
Goal: Create new health care services by
combining biomedical engineering with information
and communication technology.
“Enabling monitoring and treatment of subjects
anywhere, anytime and intervene when needed.”
Remote monitoring
–
Remote measurement of vital
biosignals
without interfering with daily activities.
Remotely supervised training & treatment
–
monitoring + feedback that enable a patient to
train when and where convenient and with the same
quality of training as in a clinical environment.
Benefits
Remote Monitoring
Less intramural care (costs)
More freedom for patient
Peace of mind
Remotely Supervised Treatment
High intensity training possible (more = better)
Training in natural environment translates to more
effective training
Puts patient in driver seat
Clinician can ‘treat’ several patients at the same time
Main challenges are technological feasibility and
clinical/patient acceptance.
Example: Tele
-
Treatment of
Chronic Back Pain
Studies report a change in activity level due to chronic
back pain.
Clinical study: 29 chronic back pain patients and 20
asymptomatic controls
Activity levels monitored for 7 consecutive days using an
3D inertial motion sensor
Overall activity levels the same but activity patterns
different between groups.
Will normalization of activity
patterns through feedback
improve outcome? (Clinical
trial running)
Example: Tele
-
Treatment of
Neck/Shoulder Pain
Chronic shoulder/neck pain typically shows no clear
physiological overloading.
Solution: Design a remote feedback system to warn
patients when insufficient relaxation occurs.
Muscle relaxation assessed via surface EMG with real
-
time feedback provided to patient and therapist.
100 patients treated in Belgium, Germany, Sweden,
and the Netherlands
Outcome as good as
classic treatment
Approach appreciated
by patients and therapists
INSTITUT FOR
SUNDHEDSVIDENSKAB OG
TEKNOLOGI
AALBORG UNIVERSITY
AALBORG, DENMARK
DEPT. OF
H
EALTH
S
CIENCE AND
T
ECHNOLOGY
TeleKat
project applies
User Driven Innovation to develop
wireless
telehomecare
technology enabling COPD patients
to perform self
-
monitoring of their status, and to maintain
rehabilitation activities in their homes.
TeleKat
COPD (KOL)
Brian Caulfield, Academic Director
•
Technology to monitor older adults
•
Systems deployed to 620 people
•
Building Predictive models based on data collected
TRIL Gait Analysis Platform (GAP) consists of:
•
Pressure sensing walkway (
Tactex
, S4 Sensors, Victoria, BC,
Canada)
•
Two
SHIMMER™ kinematic sensors worn on the subject’s shanks
•
Two orthogonally mounted web cameras
Unobtrusive capture of gait parameters and
physiological data in 600 patients.
Data used develop diagnostics capabilities to detect
increased gait variability & unsteadiness in elderly
people
(Predicting fall risk to 85% accuracy).
Can help with early identification of onset of diseases
such as Parkinson’s
Gait Analysis Platform
(
GAP)
Wellness and Exercise
A complete home/work technology platform has
been developed for the project, using a wearable
wireless sensors system (SHIMMERs™) and an open
shareable
software platform (
BioMOBIUS
™).
This facilitates effective monitoring and
biofeedback during exercise whilst enhancing end
-
user motivation and involvement in the process.
Balance & Strength Exercise
Balance and Strength Exercise (
BaSE
) program includes
console
-
based system installed in each of the
participant’s homes.
System guides the user through each of their exercises,
reminding them of correct way to execute each
movement.
System prompts participant to carry out prescribed
number of exercise repetitions.
Using a combination of camera and kinematic sensors,
BaSE
system provides real
-
time feedback to participant
on their performance and transmits data collected to
the physiotherapist.
Allows monitoring and modification of prescribed
exercise programs between clinic visits.
Overview
Personalized Models
Health Maintenance
Gaps
Summary
Gaps for Personalized Modeling
How to validate model predictions (especially
for internal quantities such as muscle, joint, and
bone loads)?
How to calibrate “unobservable” parameters to
which model predictions are sensitive?.
How to create personalized neural control
models?
How to make generation of model
-
based
predictions fast and easy for a clinical setting?
Gaps for Health Maintenance
User
-
centered development
Effective technology transfer
Demonstration of efficacy
Need for models to identify predictive
variables
Overview
Personalized Models
Health Maintenance
Gaps
Summary
Summary for Personalized Modeling
We have lots of technology! What we need are
better ways to predict how to use technology to
achieve significant improvements in mobility for
specific patients and impairments.
Personalized modeling is one option for predicting
how to use technology more effectively.
Creating personalized musculoskeletal models is
not enough
–
we also need to include
personalized neural control/
neuroplasticity
models
so that patient responses to possible treatments
can be predicted.
Summary for Health Maintenance
Technology for monitoring in progress
Collaborations world
-
wide
Need for user
-
centered development
Predictive models needed
Enter the password to open this PDF file:
File name:
-
File size:
-
Title:
-
Author:
-
Subject:
-
Keywords:
-
Creation Date:
-
Modification Date:
-
Creator:
-
PDF Producer:
-
PDF Version:
-
Page Count:
-
Preparing document for printing…
0%
Comments 0
Log in to post a comment