The 70–150 Hz detector separated movement from rest in all 13 analyzed SEEG patients, offline and in task-locked windows

This continuation of a preregistered plan applied an unchanged 70–150 Hz high-gamma movement-versus-rest detector to the next two patients (sub-07 and sub-06) in the public ds006253 SEEG cohort. Using log-transformed 0.5 s window means and shrinkage linear discriminant analysis, trained on the first 60% of time-ordered trials and tested on the last 40%, the held-out AUC was 0.851 for sub-07 and 0.787 for sub-06; each within-trial label-swap permutation test gave p = 0.0001 (10,000 permutations). Both values fell within the previous 11-patient range (0.569–0.965). Across all 13 patients with results, mean AUC was 0.824, median 0.836, and range 0.569–0.965; all 13 AUCs were above 0.5 and had permutation p < 0.05. The exact sign-flip test gave p = 0.00012. AUC did not show a detected association with channel count (50–148 channels; Spearman rho = 0.12, permutation p = 0.69). These offline results support movement detection as a possible first stage of a SEEG brain-computer interface, independent of clinical contact count. They do not demonstrate online control or distinguish movement from other task events. The analysis used one time-ordered 60/40 split and one mouse task; MOVE and REST were from different task phases, and 9 of the 22 cohort patients remained unanalyzed.

stereo-EEG · brain-computer interface · high-gamma activity · movement decoding · linear discriminant analysis · held-out AUC · permutation test