Weighted Errors-in-Variables Modelling for Detection of Cortico-Muscular Couplings

Functional cortico-muscular couplings are commonly assessed through cortico-muscular coherence (CMC) analysis, a measure of linear dependency between electroencephalogram (EEG) and electromyogram (EMG) signals. However, the presence of noise in EEG and EMG signals may exceed the strength of synchronous components, posing challenges in reliably detecting CMC. This study introduces an approach based on weighted errors-in-variables (EIV) modelling to extract relevant versions of cortical and muscular signals governing movement control from noisy EEG and EMG signals, aiming to enhance co-herence estimation. Two algorithms are presented for identifying the underlying EIV system: one employing total least squares and the other utilizing weighted total least squares, where knowledge of the unequal variance of observations is incorporated into the regression. The effectiveness of the proposed method is evaluated using synthetic and neurophysiological data, revealing substantial improvements in CMC detection.