Fig. 3

EMG SNR during hand grasp and hand extension using all 32 EMG channels (a), the top five most correlated EMG channels (b), and top five EMG channels with the highest SNR (c). (a) Across all 32 EMG channels, the SNR of the paretic EMG was significantly lower than the SNR of the nonparetic EMG and healthy EMG for hand extension. In contrast, no differences were observed among groups for hand grasp. Similarly, no within-group differences were found between hand grasp and hand extension. (b) When using only the top five most correlated EMG channels as a proxy for recording directly from the respective flexors or extensors, a similar trend is seen, although the difference in paretic extension is no longer significant. (c) When using only the top five EMG channels with the highest SNR as another proxy for recording directly from the respective flexors or extensors, the trend is still present, and extension is significantly worse for the paretic hand than the healthy control group. Box plots show the median, interquartile range, and most extreme non-outlier values. Circles denote outliers. Asterisk (*) denotes p < 0.05, pairwise rank-sum tests with correction for multiple comparisons. N = 10 control participants and 10 stroke participants (nonparetic and paretic)