Baseline EMG alters muscle synergy activations while preserving spatial structure in healthy and post-stroke subjects.

backgroundMuscle synergy analyses often exclude low-level activity by segmenting electromyography signals to task-dependent epochs, potentially discarding physiologically relevant baseline signals. We evaluated how padding type (resting baseline, shuffled baseline, Gaussian noise, and zero) and padding length (0-2T, where T represents task-dependent duration identified using a threshold method) affect upper-limb synergy extraction in healthy adults and stroke survivors performing multidirectional reaching.

methodsSynergies were identified with the non-negative matrix factorization. We compared synergy number and spatial structure across conditions, and quantified activation coefficient dynamics using onset/offset shifts, within-task shape similarity, and a Smearing Index (SI) that captures spillover into baseline segments.

resultsIncluding resting baseline slightly increased synergy number, with spatial synergies remaining highly stable across padding lengths and types. In contrast, temporal activation coefficient was sensitive: padding advanced onset, delayed offset, increased SI, and reduced within-task shape similarity relative to 0T in a length-dependent manner. Noise padding inflated dimensionality and perturbed synergy structure more than real baseline, whereas zero padding tended to simplify solutions. Synergy loss and merging were threshold-sensitive and may stem from methodological choices rather than purely pathological changes. Stroke participants showed greater sensitivity to padding and fewer synergies when the baseline was excluded.

conclusionsCollectively, the results indicate that resting baseline contains structured, low-level activations that influence temporal expression rather than spatial composition of synergies. Modest baseline inclusion (e.g., 0.5T) can capture this information without distorting core structure. These findings highlight the limitations of current synergy analysis methods and their implications for interpreting pathological motor strategies.