Characterizing the Effect of Electrode Shift & Sensor Reapplication on Common sEMG Features in Lower Limb Muscles

TL;DR

This study examines how electrode shift and reapplication affect sEMG features in lower limb muscles, revealing that frequency features are stable while time-domain features vary, impacting stroke recovery monitoring.

Contribution

It provides a detailed analysis of inter-session errors caused by electrode repositioning, highlighting the stability of frequency features and variability of time-domain features in sEMG signals.

Findings

Frequency features are resilient to electrode shifts.

Time-domain features show predictable variability.

Quantifying electrode shift can improve inter-session comparisons.

Abstract

This study investigates the impact of electrode shift and sensor reapplication on common surface electromyography (sEMG) features in lower limb muscles, factors which have, thus far, precluded clinicians from being able to attribute inter-session changes in sEMG signal properties to physiological changes in patients under the context of stroke recovery monitoring. To explore these inter-session errors, we recruited 12 healthy participants to perform a selection of isometric and dynamic exercises seen within stroke assessment sessions while instrumented with high-density sEMG (HDsEMG) arrays on the gastrocnemius medialis, tibialis anterior, semitendinosus, and tensor fascia latae. Between exercise sets, the electrode arrays were intentionally shifted and reapplied to quantify errors in signal features, using 3D scanning equipment to extract the ground truth shift performed. Results revealed that while frequency-domain features (mean, median, and peak frequency) demonstrated high resilience to the inter-session changes, the time-domain features (integrated EMG and max envelope amplitude) showed a greater, yet predictable, variability. In all, these findings suggest that should we be able to quantify placement shift, this can support direct inter-session feature comparisons, improving the reliability of sEMG-based stroke recovery assessments and offering insights for improving remote stroke rehabilitation technologies.