Contiguous Temporal Withholding for Hemispheric Analysis in Frontal EEG During Cycling

We present a methodological framework for evaluating temporal sensitivity in frontal EEG recordings during cycling. The approach departs from randomized train–test splits and instead implements contiguous temporal withholding to preserve physiological ordering. Class-wise recall is proposed as a time-indexed indicator of hemispheric recognizability under progressive temporal displacement. The method was evaluated on proof-of-concept data obtained from four healthy volunteers performing sustained lower-limb exercise while frontal EEG was recorded. Signals were segmented into contraction-aligned epochs (i.e., time-locked signal segments) and organized as within-subject temporal sequences. A contiguous 20% segment of data was withheld across 20 equally spaced temporal positions, and model training was repeated five times under identical hyperparameters. To validate temporal sensitivity, the method is evaluated on its ability to capture evolving hemispheric recognizability across the pedaling sequence, rather than separability. Interhemispheric differences are subsequently derived from these recall-based measures. We assess whether block-to-block recall modulation exceeds repetition-related variability. Results show consistent block-dependent modulation with repetition dispersion, indicating the method detects structured temporal shifts rather than stochastic training effects. The framework provides a structured way to evaluate hemispheric differentiation during gradual cortical reorganization, without relying on maximal classification accuracy. • Contiguous-block temporal withholding • Repeated-training recall estimation as a temporal stability index • Block-wise interhemispheric analysis