Characterizing sEMG Feature Errors in Lower Limb Muscles Due to Electrode Location and Skin–Electrode Interface Changes

Surface electromyography (sEMG) is popular for monitoring muscle activation. However, it remains underutilized in clinical settings such as stroke rehabilitation. This is largely due to inter-session errors from variabilities in electrode location, the skin-electrode interface, and other inter-session effects. These errors obscure the ability to attribute sEMG signal changes to physiological patient changes. In this study, we quantified different sources of error by applying high-density sEMG (HDsEMG) arrays on the gastrocnemius medialis, tibialis anterior, semitendinosus, and tensor fascia latae of 12 healthy participants performing isometric and dynamic exercises common in stroke assessments. Between exercise sets (sessions), we shifted and reapplied HDsEMG arrays to analyze three error conditions: 1) same session errors (SSE) arising from changes in the recording electrode and electrode location, 2) same location errors (SLE) from changes in recording electrodes at the same location across sessions, and 3) same electrode errors (SEE) from comparing the same recording electrode at a different location across sessions. For SSE, frequency-domain features (mean, median, peak frequency) showed <30% error at <10cm between electrodes. Time-domain features (integrated EMG, max envelope amplitude) exhibited greater errors, with the tensor fasciae latae exhibiting >50% error at 10cm between electrodes. Contrastingly, we saw <19% error across frequency- and time-domain features for SLE and SEE after correcting for differences in electrode location using SSE trends. These results show differences in electrode location were most responsible for sEMG feature errors. Thus, accounting for electrode location could improve inter-session sEMG feature comparisons, enhancing the viability of sEMG-informed physiological assessments.