Invited Review
Image-Based Musculoskeletal Modeling:
Applications, Advances, and Future Opportunities
Silvia S. Blemker, PhD,
1,5
*
Deanna S. Asakawa, PhD,
2
Garry E. Gold, MD,
4
and
Scott L. Delp, PhD
2,3
Computer models of the musculoskeletal system are
broadly used to study the mechanisms of musculoskeletal
disorders and to simulate surgical treatments. Musculo-
skeletal models have historically been created based on
data derived in anatomical and biomechanical studies of
cadaveric specimens. MRI offers an abundance of novel
methods for acquisition of data from living subjects and is
revolutionizing the field of musculoskeletal modeling. The
need to create accurate, individualized models of the mus-
culoskeletal system is driving advances in MRI techniques
including static imaging, dynamic imaging, diffusion imag-
ing, body imaging, pulse-sequence design, and coil design.
These techniques apply to imaging musculoskeletal anat-
omy, muscle architecture, joint motions, muscle moment
arms, and muscle tissue deformations. Further advance-
ments in image-based musculoskeletal modeling will ex-
pand the accuracy and utility of models used to study
musculoskeletal and neuromuscular impairments.
Key Words: skeletal muscle; biomechanics; human move-
ment; muscle architecture; magnetic resonance imaging;
dynamic imaging; musculoskeletal modeling
J. Magn. Reson. Imaging 2007;25:441– 451.
© 2007 Wiley-Liss, Inc.
THE OUTCOMES OF SURGERIES to correct disabling
movement abnormalities are unpredictable, and some-
times unsuccessful. Theoretically, patients’ abnormal
movement patterns can be improved by identifying the
biomechanical factors that contribute to abnormal
movement and designing treatments accordingly. How-
ever, many factors can contribute to the abnormal
movement. For example, persons with cerebral palsy
exhibit disturbances in voluntary control (1), muscle
spasticity (2), static muscle contractures (3), bone de-
formities that alter muscle paths (4), and limb malalign-
ment (5). Current diagnostic methods do not allow cli-
nicians to reliably differentiate between the potential
causes of abnormal movement to determine the most
appropriate treatment.
We believe that computer models of the musculoskel-
etal system can help provide a scientific basis for treat-
ing movement disorders. Models allow us to answer
“what if” questions, isolate individual sources for im-
pairment, and estimate parameters, such as muscle
forces, that are difficult to measure experimentally. In
recent years, computational models that characterize
the three-dimensional surface geometry of bones, kine-
matics of joints, and the force-generating capacity of
muscles have emerged as powerful tools for investigat-
ing muscle function. Models have been used to simu-
late orthopedic procedures, such as osteotomies (6,7),
tendon transfers (8 –11), tendon lengthenings (12,13),
and total joint replacements (14 –16). Musculoskeletal
models, combined with dynamic simulation, have been
used to understand normal (17) and pathological hu-
man movement (18).
These model-based studies have provided clinically
useful insights and general guidelines; however, the
results may have limited applicability to the treatment
of individual patients. There have been limited sources
for data that can be used to create musculoskeletal
models and to test the predictions made by simulating
treatments. The input parameters are typically based
on an accumulation of cadaveric measurements from a
range of studies. The predictions made by musculoskel-
etal models are tested with average data from unim-
paired adult populations. These traits pose two impor-
tant problems for using models to study individual
patients.
First, the models generally represent the musculo-
skeletal anatomy and function of average adult sub-
jects [e.g., Delp et al (19)]. It is not clear how muscu-
loskeletal deformities or even simply variations in size
1
Department of Mechanical & Aerospace Engineering, University of
Virginia, Charlottesville, Virginia, USA.
2
Department of Bioengineering, Stanford University, Stanford, Califor-
nia, USA.
3
Department of Mechanical Engineering, Stanford University, Stan-
ford, California, USA.
4
Department of Radiology, Stanford University, Stanford, California,
USA.
5
Department of Biomedical Engineering, University of Virginia, Char-
lottesville, Virginia, USA.
Contract grant sponsor: National Institutes of Health Roadmap for
Medical Research; Contract grant number: U54 GM072970; Contract
grant sponsor: National Institutes of Health; Contract grant numbers:
R01 HD33929, R01 HD046814, R01 EB002524, R01 EB005790.
*Address reprint requests to: S.S.B., Department of Mechanical & Aero-
space Engineering, University of Virginia, 122 Engineer’s Way, P.O. Box
400746, Charlottesville, Virginia 22904-4746.
E-mail: ssblemker@virginia.edu
Received May 4 2006; Accepted September 13 2006.
DOI 10.1002/jmri.20805
Published online in Wiley InterScience (www.interscience.wiley.com).
JOURNAL OF MAGNETIC RESONANCE IMAGING 25:441– 451 (2007)
© 2007 Wiley-Liss, Inc. 441