Clinician-labeled spiking SEEG contacts had more high-gamma bursts, with separation largely shared with broadband amplitude

In 42 NREM sleep SEEG runs from OpenNeuro datasets ds007118, ds007119, and ds007120, we compared 219 clinician-labeled spiking contacts with 664 normal contacts using 60-second middle-of-record clips. High-gamma (70–150 Hz) burst rate had a mean within-run AUC of 0.755 (95% CI 0.691–0.816), and spiking contacts had higher burst rates in 36/42 runs (median ratio about 1.4×). Line length, 15–50 Hz transient rate, and high-gamma power also separated labels; high-gamma envelope kurtosis did not. After regression on line length and 1–30 Hz power, residual AUCs fell to 0.565 for high-gamma burst rate and 0.611 for transient rate, indicating that much of the separation was shared with broadband amplitude. These contact-level clinical labels are not event-level interictal discharge marks; the analysis used monopolar signals, did not control for anatomy, and sampled only 60 seconds per run. The results support ranking contacts for review, not a hard exclusion rule. The brain-computer-interface consequence is an inference from sleep data, and no task-related high-gamma activity was available.

SEEG · high-gamma activity · epilepsy · brain-computer interfaces · interictal epileptiform discharges · signal processing