Ursula Trinler1, Fabien Leboeuf2, Kristen Hollands2, Richard Jones2, Richard Baker2. 1. University of Salford, School of Health Science, Allerton Building, Frederick Road Campus, Salford, M6 6PU, United Kingdom; BG Unfallklinik Ludwigshafen, Zentrum für Bewegungsanalytik, Forschung und Lehre, Ludwig-Guttmann Straße 13, 67071 Ludwigshafen, Germany. Electronic address: Ursula.trinler@bgu-ludwigshafen.de. 2. University of Salford, School of Health Science, Allerton Building, Frederick Road Campus, Salford, M6 6PU, United Kingdom.
Abstract
BACKGROUND: Muscle force estimation could improve clinical gait analysis by enhancing insight into causes of impairments and informing targeted treatments. However, it is not currently standard practice to use muscle force models to augment clinical gait analysis, partly, because robust validations of estimated muscle activations, underpinning force modelling processes, against recorded electromyography (EMG) are lacking. RESEARCH QUESTION: Therefore, in order to facilitate future clinical use, this study sought to validate estimated lower limb muscle activation using two mathematical models (static optimisation SO, computed muscle control CMC) against recorded muscle activations of ten healthy participants. METHODS: Participants walked at five speeds. Visual agreement in activation onset and offset as well as linear correlation (r) and mean absolute error (MAE) between models and EMG were evaluated. RESULTS: MAE between measured and recorded activations were variable across speeds (SO vs EMG 15-68%, CMC vs EMG 13-69%). Slower speeds resulted in smaller deviations (mean MAE < 30%) than faster speeds. Correlation was high (r > 0.5) for only 11/40 (CMC) and 6/40 (SO) conditions (muscles X speeds) compared to EMG. SIGNIFICANCE: Modelling approaches do not yet show sufficient consistency of agreement between estimated and recorded muscle activation to support recommending immediate clinical adoption of muscle force modelling. This may be because assumptions underlying muscle activation estimations (e.g. muscles' anatomy and maximum voluntary contraction) are not yet sufficiently individualizable. Future research needs to find timely and cost efficient ways to scale musculoskeletal models for better individualisation to facilitate future clinical implementation.
BACKGROUND: Muscle force estimation could improve clinical gait analysis by enhancing insight into causes of impairments and informing targeted treatments. However, it is not currently standard practice to use muscle force models to augment clinical gait analysis, partly, because robust validations of estimated muscle activations, underpinning force modelling processes, against recorded electromyography (EMG) are lacking. RESEARCH QUESTION: Therefore, in order to facilitate future clinical use, this study sought to validate estimated lower limb muscle activation using two mathematical models (static optimisation SO, computed muscle control CMC) against recorded muscle activations of ten healthy participants. METHODS:Participants walked at five speeds. Visual agreement in activation onset and offset as well as linear correlation (r) and mean absolute error (MAE) between models and EMG were evaluated. RESULTS: MAE between measured and recorded activations were variable across speeds (SO vs EMG 15-68%, CMC vs EMG 13-69%). Slower speeds resulted in smaller deviations (mean MAE < 30%) than faster speeds. Correlation was high (r > 0.5) for only 11/40 (CMC) and 6/40 (SO) conditions (muscles X speeds) compared to EMG. SIGNIFICANCE: Modelling approaches do not yet show sufficient consistency of agreement between estimated and recorded muscle activation to support recommending immediate clinical adoption of muscle force modelling. This may be because assumptions underlying muscle activation estimations (e.g. muscles' anatomy and maximum voluntary contraction) are not yet sufficiently individualizable. Future research needs to find timely and cost efficient ways to scale musculoskeletal models for better individualisation to facilitate future clinical implementation.
Authors: Mohammad T Karimi; Fatemeh Hemmati; Mohammad A Mardani; Keyvan Sharifmoradi; Seyed Iman Hosseini; Reza Fadayevatan; Amir Esrafilian Journal: Phys Eng Sci Med Date: 2021-02-08