StimEMG: An Electromyogram Recording System With Real-Time Removal of Time-Varying Electrical Stimulation Artifacts

• Published in: IEEE transactions on neural systems and rehabilitation engineering Vol. 33; pp. 1305 - 1315
• Main Authors: Zhao, Jiashun, Yuan, Rui, Shin, Henry, Ji, Run, Zheng, Yang
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.01.2025
• Subjects: adaptive filter; Adaptive filters; Adult; Algorithms; Band-pass filters; Computer Simulation; Computer Systems; Correlation; Dermis; Electric Stimulation - methods; Electrical stimulation; Electromyography; Electromyography - instrumentation; Electromyography - methods; EMG; Equipment Design; Female; Humans; Impedance; Iron; Male; Muscle Contraction - physiology; Muscle, Skeletal - physiology; Muscles; Real-time systems; recursive least squares; Reproducibility of Results; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Signal-To-Noise Ratio; stimulation artifact; Young Adult
• ISSN: 1534-4320; 1558-0210; 1558-0210
• DOI: 10.1109/TNSRE.2025.3555572

A closed-loop Functional Electrical Stimulation (FES) system that incorporates electromyogram (EMG) signal feedback provides more effective assistance to paralytic patients in maintaining and recovering their motor abilities. However, the closed-loop FES system with real-time adjustment of stimulation parameters tends to introduce time-varying stimulation artifacts in EMG signals, challenging the removal of stimulation artifacts that aims at more accurate monitoring of muscle contraction status. Therefore, an EMG acquisition system that embeds a stimulation artifact generation (SAG) circuit and the Recursive Least Squares (RLS) adaptive filter was developed in this study and named StimEMG. The SAG-RLS strategy was tested using the simulated contaminated EMG signals and the StimEMG system was tested in an experimental study with 8 subjects. Both the simulation and the experimental study showed that the SAG-RLS method obtained a higher correlation (R<inline-formula> <tex-math notation="LaTeX">{}^{{2}}\text {)} </tex-math></inline-formula> between the denoised EMG and the corresponding clean EMG or EMG segments compared with the current Gram-Schmidt-based (GSB) method (simulation study, <inline-formula> <tex-math notation="LaTeX">0.98\pm 0.0044 </tex-math></inline-formula> v.s. <inline-formula> <tex-math notation="LaTeX">0.65\pm 0.3217 </tex-math></inline-formula>; experimental study, <inline-formula> <tex-math notation="LaTeX">0.99\pm 0.0024 </tex-math></inline-formula> v.s. <inline-formula> <tex-math notation="LaTeX">0.52\pm 0.2105 </tex-math></inline-formula>). Meanwhile, the SAG-RLS method can suppress stimulation artifact more effectively, resulting a higher signal-to-noise ratio (simulation study: <inline-formula> <tex-math notation="LaTeX">12.83\pm 2.1745 </tex-math></inline-formula> v.s. <inline-formula> <tex-math notation="LaTeX">1.54\pm 1.3106 </tex-math></inline-formula>) and higher noise rejection ratio (experimental study:<inline-formula> <tex-math notation="LaTeX">2.32\pm 0.7046 </tex-math></inline-formula> v.s. <inline-formula> <tex-math notation="LaTeX">1.92\pm 0.8014 </tex-math></inline-formula>). The significantly improved performance is speculated to result from the ability of the SAG unit to precisely and timely capture the variation of the stimulation artifacts caused by the change of stimulation parameters, unlike previous methods relying on the stability of the characteristic of stimulation artifacts in the contaminated EMG signals. The developed StimEMG system provides a robust EMG acquisition module for the closed-loop FES system.