In 34 SEEG patients, interictal broadband and delta power are higher at seizure-onset-zone channels, but the 30–45 Hz aperiodic exponent does not separate them before spatial normalization

We asked whether spectral features from the first 60 seconds of two interictal SEEG clips per patient separate clinically labeled seizure-onset-zone (SOZ) channels from other channels. We analyzed 34 of 36 patients with features (3,708 bipolar channels; 367 SOZ channels) from OpenNeuro ds004100 v1.1.3. Raw within-patient AUCs showed higher SOZ log delta power (mean AUC 0.707; 88% of patients above 0.5; BH q<0.001), total power (0.687), and line length (0.654); relative high-gamma power was lower (0.390). The raw 30–45 Hz aperiodic exponent did not separate SOZ channels (AUC 0.492, p=0.830). A leave-one-patient-out spatial normative map based on 30 nearest non-SOZ, non-resected channels from other patients reduced band-power AUCs (delta 0.652, q<0.001) and revealed a steeper 30–45 Hz exponent at SOZ channels (AUC 0.590, p=0.015, q=0.023). Seizure-free (n=20) and not-seizure-free (n=14) groups had similar AUCs. Effects are modest and do not establish clinical localization performance. Limitations include 60 seconds rather than the full 5-minute clips, sampling rates of 256–1024 Hz across patients, clinical SOZ labels, no spike or high-frequency oscillation detection, dependence among patients and channels due to electrode sampling, exploratory tests without preregistration, unchecked fsaverage coordinate accuracy, and two HUP112 runs with no feature rows for an unexplored contact-matching reason.

SEEG · epilepsy · seizure onset zone · interictal · spectral power · aperiodic exponent · normative mapping · AUC