EMG-Driven Optimal Estimation of Subject-SPECIFIC Hill Model Muscle–Tendon Parameters of the Knee Joint Actuators

• Published in: IEEE transactions on biomedical engineering Vol. 64; no. 9; pp. 2253 - 2262
• Main Authors: Falisse, Antoine, Van Rossom, Sam, Jonkers, Ilse, De Groote, Friedl
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.09.2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Actuators; Adult; Computational modeling; Computer Simulation; Elastic Modulus - physiology; Electromyography; Electromyography - methods; Estimation; Excitation Contraction Coupling - physiology; Female; Flexors; Hill model; Humans; Identification methods; Inverse dynamics; Joints (anatomy); Knee; Knee Joint - physiology; Male; Mathematical models; Models, Neurological; Movement - physiology; Muscle Contraction - physiology; Muscle, Skeletal - physiology; Muscles; musculoskeletal modeling; Optimal control; Optimization; Parameter estimation; Parameter identification; Range of Motion, Articular - physiology; Tendons; Tendons - physiology; Tensile Strength - physiology; Torque
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2016.2630009

Objective: the purpose of this paper is to propose an optimal control problem formulation to estimate subject-specific Hill model muscle-tendon (MT-) parameters of the knee joint actuators by optimizing the fit between experimental and model-based knee moments. Additionally, this paper aims at determining which sets of functional motions contain the necessary information to identify the MT-parameters. Methods: the optimal control and parameter estimation problem underlying the MT-parameter estimation is solved for subject-specific MT-parameters via direct collocation using an electromyography-driven musculoskeletal model. The sets of motions containing sufficient information to identify the MT-parameters are determined by evaluating knee moments simulated based on subject-specific MT-parameters against experimental moments. Results: the MT-parameter estimation problem was solved in about 30 CPU minutes. MT-parameters could be identified from only seven of the 62 investigated sets of motions, underlining the importance of the experimental protocol. Using subject-specific MT-parameters instead of more common linearly scaled MT-parameters improved the fit between inverse dynamics moments and simulated moments by about 30% in terms of the coefficient of determination (from <inline-formula><tex-math notation="LaTeX">\text{0.57} \pm \text{0.20}</tex-math></inline-formula> to <inline-formula><tex-math notation="LaTeX">\text{0.74} \pm \text{0.14}</tex-math></inline-formula>) and by about 26% in terms of the root mean square error (from <inline-formula><tex-math notation="LaTeX">\text{15.98} \pm \text{6.85}</tex-math></inline-formula> to <inline-formula><tex-math notation="LaTeX">\text{11.85} \pm \text{4.12}\,{\text{N}} \cdot {\text{m}}</tex-math> </inline-formula>). In particular, subject-specific MT-parameters of the knee flexors were very different from linearly scaled MT-parameters. Conclusion: we introduced a computationally efficient optimal control problem formulation and provided guidelines for designing an experimental protocol to estimate subject-specific MT-parameters improving the accuracy of motion simulations. Significance: the proposed formulation opens new perspectives for subject-specific musculoskeletal modeling, which might be beneficial for simulating and understanding pathological motions.