In 11 SEEG patients, high-gamma separates post-onset from pre-onset movement windows

In 11 patients with stereo-electroencephalography (SEEG) from OpenNeuro ds006253, we tested whether author-provided 70–150 Hz high-gamma activity distinguishes the 0.5 s after mouse movement onset (MOVE) from the 0.5 s before onset (PRE), within the same trial. Trials were ordered by stimulus onset; decoders trained on the first 60% and tested on the last 40%, using log10 window-mean features and Ledoit–Wolf shrinkage LDA. MOVE-vs-PRE mean held-out AUC was 0.696 (range 0.548–0.955); all 11 patients were above 0.5 (exact one-sided sign-flip p=0.0005), and 10/11 had Holm-adjusted within-patient p<0.05. The MOVE-vs-REST replication mean AUC was 0.825 (0.569–0.965; 11/11 above 0.5; p=0.0005), including AUC 0.953 and p=0.0001 in the newly tested sub-21. Applied without retraining, the MOVE-vs-REST detector separated PRE from REST with mean AUC 0.734 (0.532–0.963; 11/11 above 0.5; p=0.0005). Its MOVE-vs-REST AUC exceeded its PRE-vs-REST AUC by 0.091 (10/11; two-sided p=0.003). In the median patient, 77% of above-chance MOVE AUC was present in PRE. In an exploratory comparison, the dedicated MOVE-vs-PRE decoder exceeded the MOVE-vs-REST detector on MOVE-vs-PRE by 0.070 AUC (9/11; p=0.012). These results show movement-onset-locked information, but MOVE-vs-REST performance overstates onset timing; the detector already scores activity before movement. PRE can include stimulus response and decision activity, so this analysis cannot attribute all separation to movement. The study is offline, uses one task and dataset, 0.5 s windows, and one time-ordered split; it does not test online latency or other frequency bands.

stereo-electroencephalography · high-gamma activity · movement onset · brain-computer interface · linear discriminant analysis · OpenNeuro ds006253