Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Syst Biol Med
Impact Factor: 3.542

Toward modeling locomotion using electromyography‐informed 3D models: application to cerebral palsy

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

This position paper proposes a modeling pipeline to develop clinically relevant neuromusculoskeletal models to understand and treat complex neurological disorders. Although applicable to a variety of neurological conditions, we provide direct pipeline applicative examples in the context of cerebral palsy (CP). This paper highlights technologies in: (1) patient‐specific segmental rigid body models developed from magnetic resonance imaging for use in inverse kinematics and inverse dynamics pipelines; (2) efficient population‐based approaches to derive skeletal models and muscle origins/insertions that are useful for population statistics and consistent creation of continuum models; (3) continuum muscle descriptions to account for complex muscle architecture including spatially varying material properties with muscle wrapping; (4) muscle and tendon properties specific to CP; and (5) neural‐based electromyography‐informed methods for muscle force prediction. This represents a novel modeling pipeline that couples for the first time electromyography extracted features of disrupted neuromuscular behavior with advanced numerical methods for modeling CP‐specific musculoskeletal morphology and function. The translation of such pipeline to the clinical level will provide a new class of biomarkers that objectively describe the neuromusculoskeletal determinants of pathological locomotion and complement current clinical assessment techniques, which often rely on subjective judgment. WIREs Syst Biol Med 2017, 9:e1368. doi: 10.1002/wsbm.1368 This article is categorized under: Analytical and Computational Methods > Computational Methods Models of Systems Properties and Processes > Organ, Tissue, and Physiological Models
Varying amount of subject‐specificity in relation to representing the skeletal system of children with cerebral palsy: a scaled generic model (left), a scaled generic model including some specific deformations, for example, 30 degrees tibial torsion (center), and a subject‐specific model obtained from medical images (right).
[ Normal View | Magnified View ]
Closed‐loop neural‐based musculoskeletal modeling formulation. It comprises two main components: the neural‐based forward dynamics musculoskeletal model and the static optimization component. Recording and modeling uncertainties typically limit forward dynamics prediction capacity. The static optimization component modulates the input excitations to account for forward dynamics prediction errors, that is, see for example gluteus maximus excitation and hip flexion‐extension moment prediction error in this figure. In a subject‐specific musculoskeletal model, the transfer function between the level of excitation adjustment and moment prediction error is L‐shaped, that is, all dynamically consistent neural solutions exist in a neighborhood of the experimental input data.
[ Normal View | Magnified View ]
(a) High‐density electromyograms (EMGs) are recorded from a tremor patient extensor carpi radialis during a resting task. EMG decomposition discerns the underlying motor unit firings, which corresponds to those produced by spinal motor neurons. In the tremor condition, motor units fire with an abnormally high level of synchronization, that is, see right‐hand histogram depicting high probability for any two given motor units of firing within negligible time delays. (b) Five‐dimensional synergy structure extracted from 16 lower limb muscles EMGs across four different locomotion tasks. The rhythmicity of the extracted primitives may reflect the dynamics of spinal pattern generators active during cyclic tasks. (a and b) Advanced recording and processing techniques give access to the function of neural microstructures (i.e., motor neurons) and macrostructures (i.e., muscle synergies) in vivo in the intact moving human.
[ Normal View | Magnified View ]
In the execution of a motor task, synaptic inputs are produced as a result of neural microstructures (i.e., neurons) as well as macrostructures reflecting the coordinated interaction between populations of motor, sensory, and inter‐neurons in the generation of rhythmic motor programs in the spinal cord (i.e., neural synergies). Synaptic inputs ultimately converge to pools of motor neurons, which integrate neural information and transform it into output spike trains (i.e., neural drive) sent to innervated muscles. Neurally controlled muscle–tendon units translate the neural drive into mechanical viscoelastic forces, thus causing articular and segmental body accelerations, multijoint coordination and compliant body interaction with the environment.
[ Normal View | Magnified View ]
(a) Magnetic resonance imaging scan of both lower legs (L, left; R, right) of a child with spastic‐type cerebral palsy (SCP, right hemiplegia). Medial (MG) and lateral (LG) gastrocnemius muscles are segmented to measure muscle volume. (b) Active muscle properties, ankle torque versus angle, of individuals with SCP and typically developing (TD) peers (mean ± SEM). (Reprinted with permission from Ref . Copyright 2012 Elsevier). (c) Passive muscle properties [I], ankle torque versus angle, [II] ankle torque versus MG fascicle length of individuals with SCP and TD peers (mean ± SEM). (Reprinted with permission from Ref . Copyright 2011 Elsevier) (d) LG muscle–tendon unit and fascicle functional during one stride of walking gait in individuals with CP and TD peers, 0% stride = initial foot contact.
[ Normal View | Magnified View ]
Top: (a) semimembranosus and (b) semitendinosus muscles embedded inside hybrid mechanics host volume meshes. (c–e) Finite elastic mechanics simulations of muscle deformation, sliding, and wrapping through 45 degrees of hip flexion; and (f) muscle arc‐length changes through muscle centroid (red). Bottom: soft tissue muscle deformation during walking was derived from the deformation of a surrounding skin mesh based on inverse kinematics, that is, segmental kinematics. The predicted deformations of five muscles were validated using magnetic resonance imaging data of the same subject in two different lower limb positions.
[ Normal View | Magnified View ]
(a) Complete lower limb skin host‐mesh with embedded muscles deformed using motion capture surface markers. (b) Initial embedded muscle pose, and (c) resulting deformed muscle pose after 30 degrees flexion. Highlighted is the bipennate fiber field fitted for the rectus femoris muscle derived from cadaveric images.
[ Normal View | Magnified View ]
The Musculoskeletal Atlas Project (MAP) provides workflows for generating musculoskeletal models from clinical data using statistical models. The MAP Client software manages plugins that perform steps such as model registration and mesh fitting. These steps can use data from the MAP Database for typically developed or cerebral palsy populations (cerebral palsy limbs shown with significantly small muscles highlighted according to Z‐score). The generated models bridge the gap between rigid‐body and continuum models.
[ Normal View | Magnified View ]
Example of joint angles, joint moments, and joint powers obtained using the conventional model (Plug‐in‐Gait, in red) and a fully subject‐specific kinematic and kinetic model (in black). Results are presented only for the sagittal plane.
[ Normal View | Magnified View ]

Browse by Topic

Analytical and Computational Methods > Computational Methods
Models of Systems Properties and Processes > Organ, Tissue, and Physiological Models

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts