Interictal SEEG band power localizes the clinical seizure-onset zone; localization does not differ by surgical outcome (HUP ds004100)

Question: do simple interictal spectral features of SEEG contacts separate clinically marked seizure-onset-zone (SOZ) contacts from other contacts in the same patient, and is this separation larger in patients who became seizure-free (Engel I)? Data: OpenNeuro ds004100 v1.1.3 (HUP), all 37 SEEG patients, first 40 s of interictal run-01 (HTTP range read), with run-02 as replication (36 patients). Method: pure-numpy pipeline, CAR and bipolar montages, median Welch PSD, 16 features, within-patient AUC, sign-flip and label permutation tests, BH correction, leave-one-patient-out logistic model. Results: log delta power gave mean AUC 0.730 (CAR) and 0.705 (bipolar), p<1e-4, with AUC>0.5 in 83% of patients; run-02 gave 0.729 and 0.708. Absolute power at all bands was higher and relative alpha/beta power was lower in SOZ. The 30-45 Hz aperiodic exponent was weak and not significant after correction (CAR 0.429, bipolar 0.478). A 16-feature leave-one-patient-out model gave median AUC 0.81 (CAR). No feature separated SOZ better in seizure-free (n=20) than non-seizure-free (n=16) patients (all q>=0.94; replicated in run-02). Limits: 40 s per clip, no sleep/wake or anatomical (gray/white, mesial temporal) control, clinical SOZ is the reference, low power for the outcome contrast.

stereo-EEG · SEEG · epilepsy · seizure-onset zone · interictal · spectral power · aperiodic exponent · surgical outcome