Wearable mechanomyography sensory advancements for robust prediction of human motor intention: from multi-channel technology to data-driven computational modelling
File(s)
Author(s)
Mancero Castillo, Carlos Sebastian
Type
Thesis
Abstract
Analyzing, predicting, and decoding electrical and mechanical responses of the human muscular system in reaction to neural drive have attracted a great deal of interest in both medical and bioengineering fields. The common goal is to understand (and possibly model) the neurophysiological behaviour and to translate that into actionable functions. Mechanomyography (MMG) is a myographic signal modality which records the mechanical activity of the muscle fibres during a contraction. This modality has been investigated in the past specifically thanks to the benefits that it can propose in comparison with other modalities, such as electromyography (EMG). Specifically,
the non-invasive mechanical nature of MMG makes it suitable to tackle the limitations of more established methods which rely on electrical activity. However, despite the existing literature, there are several unknown factors affecting the usability of mechanomyography signals for clinical and non-clinical applications.
This thesis introduces novel wearable MMG sensing technology that sheds light on the functionality, modelling, and information context of this neural interfacing modality.
The first part of this thesis introduces a new mode for MMG signal capture by the modulation of normal pressure at the sensor location, enabling a systematic evaluation of different levels of contact force between the sensor and the skin. Seven able-bodied and three amputee participants took place in an experiment that consisted of the analysis of acoustic MMG activity under different levels of contact force for the classification of hand gestures. The muscles of interest for this analysis were the flexor carpi radialis (FCR), brachioradialis (BRD), extensor digitorum communis (EDC), and flexor carpi ulnaris (FCU). Results of this study indicate that increasing levels of normal force between the sensor and the skin can improve the discriminative power content of the MMG signal, showing user-specific patterns of improved classification at different levels of contact force.
These user-specific results suggest an influence of physiological and anthropomorphic factors on the properties of the MMG signal. In addition, these results provide further evidence on the effect of the level of contact force between the sensor and the skin and not only demonstrate the importance of sensor-to-skin attachment but also provide a better notion of the parameters to be considered when analysing MMG activity.
A second study was conducted to investigate the feasibility of the MMG modality in the analysis of functional muscle networks. A functional muscle network is a concept developed in the last decade which describes the functional organization of groups of muscles that play a role in the execution
of a motor task. In this study, mechanomyography is introduced as a potential tool to evaluate functional synchronization and muscle connectivity based on the mechanical response of the muscles. The analysis encompasses the study of the same muscles analysed previously in the evaluation of contact force vs acoustic MMG. Ten able-bodied and three amputee participants took part in this study, which consisted of the execution of four sustained hand gesture tasks. MMG muscle activity acquired from all four muscles was used to perform functional muscle connectivity analysis at multiple frequency bands to evaluate the topological characteristics of MMG information at low
(i.e., < 5Hz) and high (i.e., > 12 Hz) activation frequencies. The resulting muscle networks showed distinct topological signatures and specific significant differences (p-value<0.0028) in the network characteristics describing the functional segregation of the networks (i.e., clustering coefficient) across gestures and frequencies. The results also showed evident topological differences between demographics, amputees and able-bodied. These results show the potential of MMG as a tool to evaluate functional muscle connectivity and provide key insights into the neural circuitry involved
in motor coordination. The findings demonstrate the feasibility of MMG for the mapping of synergistic functional synchronization of upper-limb muscles and provide a basis for the translation of
the analysis into clinical applications.
A third study was conducted in this thesis to exploit the multi-directionality property of MMG and evaluate muscle connectivity networks in the context of a multi-dimensional space based on muscle fibre alignment. The study not only explores the usability of MMG to generate multi-dimensional muscle networks but also examines the use of intermuscular coherence for human intention detection based on the multi-directionality space of accelerometer MMG and the effect that contact force variations have on the spectrotemporal properties of this MMG modality. In this study, for the first time, three-dimensional muscle networks are proposed based on orthogonal vibrational mechanomyography (vMMG) to evaluate the functional co-modulation of muscles across multiple frequency bands and along the longitudinal, lateral, and transverse directions of the muscle fibres.
MMG activity was collected from twenty subjects using a custom-made armband of accelerometer sensors while participants performed four hand gestures. Target muscles and gestures were similar to those selected in the previous studies. IMC from all superficial muscles was decomposed into the constituent frequency bands of MMG (1-5 Hz, 5-12 Hz, 12-40 Hz). Muscle connectivity networks were generated for all frequency bands for each gesture along the three anatomical directions of muscle fibre alignment. Results show distinct topographical differences across frequency bands, gestures, and directions, and provide further evidence to support the use of vMMG in muscle connectivity analysis. This study also shows the feasibility of using vMMG-based muscle networks to compose a feature space for the classification of multiple hand gestures. To evaluate the translational capacity of the proposed technique, an analysis is also conducted on the effect of contact force, considering the changes in quality and discriminative power of vMMG. No statistical evidence was observed when a higher level of contact force was applied, supporting the robustness of the vMMG muscle networks to variations in sensor placement and tissue compression.
Through the above-mentioned three studies conducted in this thesis, it can be concluded that MMG not only offers a solution for those applications where EMG is not feasible, but it also shows to be a potential tool for clinical and non-clinical applications. The results of these studies showing limitations and the potential of the myographic signal provide the opportunity to improve the design and
applicability of the technology and enable the use of the signal in new areas where MMG has not yet been applied. Results presented in this thesis show the effect that attachment of the sensors to the skin has on both modalities of the signal (i.e., accelerometer and microphone) and guide future studies en route to exploit the properties of the signal as much as is viable based on the desired application. The analysis of functional muscle connectivity networks opens a new area of use for MMG to expand the analysis of muscle synergies and further apply the analysis and technique in the context of prosthetics, rehabilitation, and assistive devices.
the non-invasive mechanical nature of MMG makes it suitable to tackle the limitations of more established methods which rely on electrical activity. However, despite the existing literature, there are several unknown factors affecting the usability of mechanomyography signals for clinical and non-clinical applications.
This thesis introduces novel wearable MMG sensing technology that sheds light on the functionality, modelling, and information context of this neural interfacing modality.
The first part of this thesis introduces a new mode for MMG signal capture by the modulation of normal pressure at the sensor location, enabling a systematic evaluation of different levels of contact force between the sensor and the skin. Seven able-bodied and three amputee participants took place in an experiment that consisted of the analysis of acoustic MMG activity under different levels of contact force for the classification of hand gestures. The muscles of interest for this analysis were the flexor carpi radialis (FCR), brachioradialis (BRD), extensor digitorum communis (EDC), and flexor carpi ulnaris (FCU). Results of this study indicate that increasing levels of normal force between the sensor and the skin can improve the discriminative power content of the MMG signal, showing user-specific patterns of improved classification at different levels of contact force.
These user-specific results suggest an influence of physiological and anthropomorphic factors on the properties of the MMG signal. In addition, these results provide further evidence on the effect of the level of contact force between the sensor and the skin and not only demonstrate the importance of sensor-to-skin attachment but also provide a better notion of the parameters to be considered when analysing MMG activity.
A second study was conducted to investigate the feasibility of the MMG modality in the analysis of functional muscle networks. A functional muscle network is a concept developed in the last decade which describes the functional organization of groups of muscles that play a role in the execution
of a motor task. In this study, mechanomyography is introduced as a potential tool to evaluate functional synchronization and muscle connectivity based on the mechanical response of the muscles. The analysis encompasses the study of the same muscles analysed previously in the evaluation of contact force vs acoustic MMG. Ten able-bodied and three amputee participants took part in this study, which consisted of the execution of four sustained hand gesture tasks. MMG muscle activity acquired from all four muscles was used to perform functional muscle connectivity analysis at multiple frequency bands to evaluate the topological characteristics of MMG information at low
(i.e., < 5Hz) and high (i.e., > 12 Hz) activation frequencies. The resulting muscle networks showed distinct topological signatures and specific significant differences (p-value<0.0028) in the network characteristics describing the functional segregation of the networks (i.e., clustering coefficient) across gestures and frequencies. The results also showed evident topological differences between demographics, amputees and able-bodied. These results show the potential of MMG as a tool to evaluate functional muscle connectivity and provide key insights into the neural circuitry involved
in motor coordination. The findings demonstrate the feasibility of MMG for the mapping of synergistic functional synchronization of upper-limb muscles and provide a basis for the translation of
the analysis into clinical applications.
A third study was conducted in this thesis to exploit the multi-directionality property of MMG and evaluate muscle connectivity networks in the context of a multi-dimensional space based on muscle fibre alignment. The study not only explores the usability of MMG to generate multi-dimensional muscle networks but also examines the use of intermuscular coherence for human intention detection based on the multi-directionality space of accelerometer MMG and the effect that contact force variations have on the spectrotemporal properties of this MMG modality. In this study, for the first time, three-dimensional muscle networks are proposed based on orthogonal vibrational mechanomyography (vMMG) to evaluate the functional co-modulation of muscles across multiple frequency bands and along the longitudinal, lateral, and transverse directions of the muscle fibres.
MMG activity was collected from twenty subjects using a custom-made armband of accelerometer sensors while participants performed four hand gestures. Target muscles and gestures were similar to those selected in the previous studies. IMC from all superficial muscles was decomposed into the constituent frequency bands of MMG (1-5 Hz, 5-12 Hz, 12-40 Hz). Muscle connectivity networks were generated for all frequency bands for each gesture along the three anatomical directions of muscle fibre alignment. Results show distinct topographical differences across frequency bands, gestures, and directions, and provide further evidence to support the use of vMMG in muscle connectivity analysis. This study also shows the feasibility of using vMMG-based muscle networks to compose a feature space for the classification of multiple hand gestures. To evaluate the translational capacity of the proposed technique, an analysis is also conducted on the effect of contact force, considering the changes in quality and discriminative power of vMMG. No statistical evidence was observed when a higher level of contact force was applied, supporting the robustness of the vMMG muscle networks to variations in sensor placement and tissue compression.
Through the above-mentioned three studies conducted in this thesis, it can be concluded that MMG not only offers a solution for those applications where EMG is not feasible, but it also shows to be a potential tool for clinical and non-clinical applications. The results of these studies showing limitations and the potential of the myographic signal provide the opportunity to improve the design and
applicability of the technology and enable the use of the signal in new areas where MMG has not yet been applied. Results presented in this thesis show the effect that attachment of the sensors to the skin has on both modalities of the signal (i.e., accelerometer and microphone) and guide future studies en route to exploit the properties of the signal as much as is viable based on the desired application. The analysis of functional muscle connectivity networks opens a new area of use for MMG to expand the analysis of muscle synergies and further apply the analysis and technique in the context of prosthetics, rehabilitation, and assistive devices.
Version
Open Access
Date Issued
2022-10
Date Awarded
2022-12
Copyright Statement
Creative Commons Attribution NonCommercial NoDerivatives Licence
Advisor
Vaidyanathan, Ravi
Burdet, Etienne
Nandi, Dipankar
Publisher Department
Mechanical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)