High-gamma power added little to distance-based prediction of expert N1 labels in eight SEEG patients

We asked whether stimulation-evoked high-gamma (HG; 70–145 Hz) power at a SEEG contact marks expert-checked N1 CCEP labels and adds predictive information beyond distance. We analyzed the first clinical SPES run for eight STReEF patients (STREEF06–13) in OpenNeuro ds005448 v1.0.0: eight stimulated pairs per patient, both pulse polarities, 749 pulses, and 5,608 pair-contact combinations (613 N1+). Distance alone predicted the N1 label with median held-out AUC 0.845 (range 0.65–0.96). Adding HG changed median AUC by −0.002 to −0.001 in single windows and −0.003 across windows (p ≥ 0.74); only one patient gained (+0.11). At matched distance, N1+ contacts had +1.11 dB total early HG (7/8 patients, p = 0.016), while induced early HG was +0.33 dB (7/8, p = 0.15), indicating that most early excess was phase-locked to the pulse. There was no late HG suppression: significant suppression occurred in 3.4% of N1+ and 3.5% of N1− combinations (p = 0.46), near the 2.5% false-positive rate. Late HG increase was more frequent at N1+ contacts (median 19% vs 9%, 8/8, p = 0.008; not distance-matched). This study is limited to eight patients at one centre, with 10–30 pulses per pair; residual artifact may affect HG before about 50 ms, the mid window followed inspection of one patient, per-combination tests were uncorrected, and interictal epileptiform discharges were not removed.

SEEG · CCEP · N1 · high-gamma · electrical stimulation · epilepsy · brain-computer interface · OpenNeuro