Interictal amplitude features identify clinical seizure-onset channels in 25 HUP SEEG patients, including within shafts, without outcome differences

We tested 16 simple features from the first 150 s of one interictal stereoelectroencephalography (SEEG) clip per patient for separation of clinically marked seizure-onset-zone (SOZ) channels in 25 patients from the single-center HUP iEEG dataset. Median per-patient area under the ROC curve (AUC) was 0.69–0.76 for amplitude features; log delta power had median AUC 0.76 (88% of patients above 0.5; Holm-adjusted p=0.0015), and spike rate had AUC 0.70 (p=0.011). Relative high-gamma power was lower in SOZ channels (AUC 0.33, p=0.011). In within-shaft comparisons of 21 patients, line length and log total power remained significant (AUC 0.79 for each), while spike rate did not (AUC 0.61). A leave-one-patient-out logistic model had median AUC 0.81 (IQR 0.69–0.89). No feature or multivariate model differed significantly by surgical outcome. The small sample (25 patients, including 8 with poor outcomes), single center, one short clip per patient, unvalidated simple spike detector, absent sleep/wake and gray/white-matter controls, selection of only 25 of 38 SEEG patients due to download limits, and imperfect clinical SOZ labels limit interpretation. These findings describe associations with clinical labels and do not establish ground-truth localization.

stereoelectroencephalography · epilepsy · seizure-onset zone · interictal EEG · signal features · area under the curve · surgical outcome