Biomedical Engineering Reference
In-Depth Information
Same as in clinical gait analysis, research onmusculoskeletal models in the clinical
context is used not only to better understand a musculoskeletal pathology but also to
help clinicians to improve procedures, treatment and rehabilitation techniques. For
example, musculoskeletal models have been used to study stroke, spinal cord injury,
osteoarthritis and neurological deficits such as cerebral palsy [
10
,
11
].
At present, musculoskeletal simulation can be performed based on a generic or
on a subject specific model. In the first most commonly applied case, an anthropo-
metric model of the bones, muscles and tendons is created based on biomechanical
studies of cadaveric specimens [
12
,
13
] and the musculoskeletal geometries of any
average adult healthy subject [
12
-
14
]. Subsequently, this generic model is scaled
to approximate the anatomy of a particular subject. To determine the scaling factor
motion capture is used. Kinematic and kinetic parameters are obtained for each joint
during a specific motion (e.g. walking, running or jumping). The length and posi-
tion of each body segment is known and used to estimate the scaling factor needed
for a specific individual. In the case of subject specific musculoskeletal modeling,
the model is created based on imaging-data which allows the use of actual subject's
bone and muscle geometries [
15
,
16
]. However, this modeling approach still requires
assumptions and data from cadaveric specimens. In addition, it is still extremely time
consuming despite interpolation techniques used to reduce effort and time required
for the creation of this type of models.
In both cases, the inputs required to build a musculoskeletal model are (1) three-
dimensional bone surface geometry, (2) the equations of motion of the body (math-
ematical descriptions of joint kinematics, (3) parameters describing each muscle's
path geometry (defined for a range of joint motions) and (4) muscle architecture
(which defines the force-generating capacity of each muscle).
Bone and muscle surface geometries can be obtained by reconstructing muscle
and bone surfaces frommultiple series of MRI images. This is achieved by a process
known as segmentation which consists of identifying and outlining the anatomical
structures of interest in each MRI image. Typically, images from multiple imaging
series are combined to create a full limb model. At present, much research is dedi-
cated to speed up this process by developing novel algorithms to make segmentation
automatic or semi-automatic [
17
-
19
]. However, these algorithms are more success-
ful for bone than for muscle segmentation, since the boundaries between muscles do
not appear sharply in MRI images. Thus a great amount of manual segmentation is
still necessary to delineate the muscles boundaries. Depending on the purpose of the
model being created, tendons can also be traced from MRI images.
Once the muscle, tendon and bone geometries have been created, joint centers
need to be calculated and used as the center point for the moment force generated
by the muscles [
20
]. Subsequently, the kinematic structure of each joint must be
created to drive skeletal motion and finally, boundary conditions are specified to
attach muscle and tendon to bone [
21
]. In other words, a mathematical method (most
commonly the Newton-Euler method; [
22
]) is used to obtain a series of equations that
describe the translational and rotational movements of the body segments. Once the
