EEG-Based Characterization of Cortical Dynamics during Graded Resistance Exercise.

Understanding how brain activity varies during voluntary movement under different resistance conditions may provide insights into neural responses associated with physical demands and inform the development of neurorehabilitation strategies. In this study, we investigated changes in neural activity associated with increasing resistance levels during unilateral bicep curl movement. Ten healthy participants performed bicep curls under five resistance conditions while their cortical activity was recorded using electroencephalography (EEG). Movement-related cortical potentials (MRCPs), power spectral density (PSD) topographies, nonparametric spatiotemporal cluster-based permutation tests and linear mixed-effects models (LMMs) were used to characterize EEG differences across resistance conditions. Differences in MR-CPs and frequency-specific power spectral density (PSD) patterns were observed across resistance conditions, with significant spatiotemporal clusters identified during pre-movement, peak contraction, and movement execution phases. A Triplet Temporal Convolutional Network (TTCN) was also employed to learn latent representations from these EEG signals, and its discriminative capability was evaluated using a k-nearest neighbors (KNN) classifier. The learned embeddings showed separation between the resistance conditions, with classification errors occurring primarily between neighboring conditions. Using a within subject leave-one-trial-out (LOTO) evaluation strategy, the TTCN-KNN framework achieved accuracy of 97.2%. Furthermore, subject-independent evaluation were conducted using a leave-one-subject-out (LOSO) approach, achieving a maximum accuracy of 93.0%. These findings suggest that EEG signals contain discriminative information for distinguishing the resistance conditions in both within-subject and subject-independent settings, providing preliminary in sights for future EEG-based monitoring of motor effort and adaptive brain-machine interface applications.