In 34 HUP SEEG patients, interictal power, line length, and transient rate identified clinical SOZ contacts; montage differences were not significant

We assessed whether simple features of the first 60 s of one interictal SEEG clip per patient distinguish clinically labeled seizure onset zone (SOZ) contacts, whether results differ across monopolar, common average reference (CAR), and bipolar montages, and whether resected-contact AUC differs by Engel outcome. We used 35 downloaded subjects from 38 SEEG subjects in OpenNeuro ds004100 v1.1.3; one downloaded subject had no SOZ contacts among good contacts, giving n = 34 for SOZ tests. FFT preprocessing applied a 0.5 Hz high-pass and removed ±1 Hz around 60 Hz harmonics. Features included log total power, relative band powers, aperiodic slope, line length, and a simple spike-like transient rate. Within-patient AUCs were tested by sign-flip permutation with Bonferroni correction. Mean SOZ AUCs were 0.71–0.73 for log total power, 0.66–0.71 for transient rate, and 0.62–0.68 for line length; each feature was significant in all three montages after correction. Aperiodic slope was more negative in SOZ contacts (AUC 0.32–0.35) and significant in all montages. No paired montage contrast survived correction; the largest mean change was +0.061 AUC for line length (bipolar minus CAR; p = 0.0026 uncorrected, 0.077 corrected). Resected-contact AUC did not differ between Engel I (n = 18) and Engel II–IV (n = 16) for any feature or montage; the smallest uncorrected permutation p was 0.052. These results are limited to 60 s from one interictal clip, use clinical SOZ labels that are not independent ground truth, include no anatomical normalization, and rely on an unvalidated transient detector. Three of 38 SEEG subjects were unavailable, and the outcome comparison had low power.

SEEG · epilepsy · seizure onset zone · interictal · intracranial EEG · AUC · montage · HUP · OpenNeuro