A Musculoskeletal Model of the Hand and Wrist Capable of Simulating Functional Tasks

Abstract

Objective: The purpose of this work was to develop an open-source musculoskeletal model of the hand and wrist and to evaluate its performance during simulations of functional tasks.

Methods: The current model was developed by adapting and expanding upon existing models. An optimal control theory framework that combines forward-dynamics simulations with a simulated-annealing optimization was used to simulate maximum grip and pinch force. Active and passive hand opening were simulated to evaluate coordinated kinematic hand movements.

Results: The model's maximum grip force production matched experimental measures of grip force, force distribution amongst the digits, and displayed sensitivity to wrist flexion. Simulated lateral pinch strength replicated in vivo palmar pinch strength data. Additionally, predicted activations for 7 of 8 muscles fell within variability of EMG data during palmar pinch. The active and passive hand opening simulations predicted reasonable activations and demonstrated passive motion mimicking tenodesis, respectively.

Conclusion: This work advances simulation capabilities of hand and wrist models and provides a foundation for future work to build upon.

Significance: This is the first open-source musculoskeletal model of the hand and wrist to be implemented during both functional kinetic and kinematic tasks. We provide a novel simulation framework to predict maximal grip and pinch force which can be used to evaluate how potential surgical and rehabilitation interventions influence these functional outcomes while requiring minimal experimental data.

Fig. 1.

A) Our model’s representation of grasping a standard dynamometer on setting II. Contact bodies (teal cylinders) are overlaid on bone geometries. Only contact from the proximal and intermediate phalanges contribute to grip force to mimic a person holding the dynamometer (B).

Fig. 2.

Kinematic motion of sign language letter “O”.

Fig. 3.

A) Average and standard deviation of experimentally measured maximum grip force, as reported from several studies (blue, green, and purple bars) and the grip force from the 5 simulations with the best objective function values in ‘self-selected’ wrist posture (orange bar). B) Average and standard deviation of the distribution of force amongst the individual digits from the same studies and simulations; of these studies, only Kong et al. (2011) reported the distribution of force amongst the digits. C) Average and standard deviation of muscle activations from the same studies and simulations; of these studies, only Mogk and Keir (2003) report EMG signals measured during grasping. D) Illustrations of the wrist posture adopted during grip force, as reported by each experimental study and the wrist posture used for our simulation results. The overlaid color-coded vectors represent the magnitude and orientation of the initial wrist posture in a Cartesian coordinate system with the positive x-axis representing wrist extension and the positive y-axis representing radial deviation. The origin of this coordinate system is aligned with the base of the lunate.

Fig. 4.

Simulated maximum grip force was compared to experimental measures in several wrist postures. Blue, green, and purple bars represent the experimental grip force reported in [62], [63], and [64] respectively. Error bars represent the standard deviation of the experimental data. Orange bars represented the simulated maximum grip force; average and standard deviation of the 5 simulations with the best objective function values (top 10%) are displayed.

Fig. 5.

Simulated activations for the 5 pinch simulations with the best objective function values (top 10%) (orange boxplot) were compared to the range of normalized EMG data during palmar pinch (blue boxplot) [9].

Fig. 6.

Simulated activations (green line) for the hand opening task generally fell within 2 standard deviations of experimental data (1 standard deviation: black region, 2 standard deviations: light grey region). OPP and FDP did not have EMG data available for comparison.

Fig. 7.

The passive simulation displayed the coupled motion between the wrist and fingers mimicking tenodesis grasp and release. The top panel displays the prescribed wrist flexion/extension. In order from top to bottom, the remaining panels display MCP, PIP, and DIP flexion/extension of the individual digits. Note differences in y-axis ranges.

Fig. 8.

Sensitivity of grip strength simulations to tendon slack length. Adjusting the tendon slack lengths to be at optimal length in the ‘self-selected’ posture did not increase the sensitivity of simulate grip strength to wrist posture. Blue squares: experimental results from [62]. Orange circles: original simulation results from Fig. 4. Grey triangles: adjusted model with original muscle activation patterns. Black triangle: adjusted model with re-optimized muscle activation patterns.