Novel cross correlation technique allows crosstalk resistant reflex detection from surface EMG

Abstract

Existing methods for withdrawal reflex detection from surface electromyography (sEMG) do not consider the potential presence of electrical crosstalk, which in practical applications may entail reduced detection accuracy. This study estimated muscle fiber conduction velocities (CV) for the tibialis anterior (TA) and soleus (SOL) muscles of both genuine reflexes and identified crosstalk, measured during antagonistic reflex responses. These estimations were used to develop and assess a novel method for reflex detection resistant to crosstalk. Cross correlations of two single differential (SD) sEMG signals recorded along the muscle fibers were performed and two features were extracted from the resulting correlograms (average CV and maximal cross correlation). Reflex detection based on evaluation of the extracted features was compared to a conventional reflex detection method (thresholding of interval peak z-scores), applied on both SD and double differential (DD) sEMG. Intramuscular electromyography (iEMG) was used as validation for reflex detection. Apparent CV due to electrical crosstalk alone were more than one order of magnitude higher than CV estimated for genuine reflexes. Conventional reflex detection showed excellent sensitivity but poor specificity (0.19-0.76) due to the presence of crosstalk. In contrast, cross correlation analysis allowed reflex detection with significantly improved specificity (0.91-0.97). The developed methodology may be readily implemented for more reliable reflex detection.