Motor imagery (MI) recruits motor networks without overt movement and underpins many brain-computer interface (BCI) paradigms, yet how neural activity is organized across distinct task stages remains unclear. Using stereoelectroencephalography (sEEG) from ten epilepsy patients performing cued limb MI, we compared preparation and imagery stages and quantified trial-wise power changes in low-frequency (8-30 Hz) and high-frequency (60-115 Hz) bands across cortical and subcortical contacts. Low-frequency activity predominantly showed suppression during preparation followed by activation during imagery, consistent with ERD/ERS-like dynamics, whereas high-frequency responses were stronger, observed across a greater number of regions, and additionally showed activation-suppression sequences in a subset of contacts. These findings indicate that neural responses evolve differently across preparation and imagery, reflecting frequency- and region-specific dynamics rather than a uniform task-related response. Modulation in deep structures, including hippocampal subfields, suggests that MI can engage a distributed network beyond canonical sensorimotor areas. These results refine the temporal and spectral characterization of MI and may inform stage-aware BCI feature design and neurorehabilitation.
Fu et al. (Fri,) studied this question.