Comparison of filtering methods for real-time extraction of the volitional EMG component in electrically stimulated muscles

Objective Recorded electromyograms (EMG) of electrically stimulated muscles can contain both an exogenous-evoked potential (M-wave) and an endogenous, or volitional, component. This study evaluated the effectiveness of three filtering methods (i.e., high-pass, adaptive, and comb), commonly used in n...

Descripción completa

Detalles Bibliográficos
Autores: Hambly, Matthew, Sousa, Ana Carolina Cardoso de|||0000-0003-1668-8949, Pizzolato, Claudio
Tipo de recurso: artículo
Fecha de publicación:2023
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/394943
Acceso en línea:https://hdl.handle.net/2117/394943
https://dx.doi.org/10.1016/j.bspc.2023.105471
Access Level:acceso abierto
Palabra clave:Electromyography
Functional electrical stimulation (FES)
M-wave
Signal processing
Volitional electromyography (EMG)
EMG modelling
Motor unit recruitment
Electromiografia
Àrees temàtiques de la UPC::Enginyeria biomèdica
Descripción
Sumario:Objective Recorded electromyograms (EMG) of electrically stimulated muscles can contain both an exogenous-evoked potential (M-wave) and an endogenous, or volitional, component. This study evaluated the effectiveness of three filtering methods (i.e., high-pass, adaptive, and comb), commonly used in neurorehabilitation, in extracting the volitional component of simulated and experimental EMG during upper-limb tasks. Methods Volitional EMG and M-wave were simulated through a physiological model of muscle recruitment, comprising of a motor neuron pool and associated muscle fibres, superimposed to a stimulation artefact. Experimental EMG data during different levels of volitional muscle contraction in isometric and dynamic tasks were recorded from five unimpaired individuals. Electrical stimulation artefact was removed with different techniques (i.e., none, removing samples, blanking, and interpolation) to assess filter performance across time and frequency domains, and information content (i.e., Kolmogorov-Smirnov D-value). Results The experimental results agreed with the simulations, wherein the adaptive filter outperformed the other filters when using no artefact removal or removing artefact samples from the signal, while for the blanking and interpolation artefact removal methods, the adaptive and comb filters outperformed the high-pass filter. Conclusion The adaptive and comb filters best estimated volitional muscle activity in electrically stimulated muscles. Significance Results from this study will enable the enhanced design of real-time neuroprosthesis control.