In HUP SEEG, ictal markers identify clinician SOZ contacts, but their power advantage does not replicate

We evaluated automated ictal markers for localizing clinician-marked seizure-onset-zone (SOZ) contacts, their stability between seizures, and their relation to surgical outcome in OpenNeuro ds004100 (HUP), using ictal clips from 36 SEEG patients. For seizure 1, 35 patients had both SOZ and non-SOZ pairs. Mean within-patient SOZ AUCs were 0.70 for the Epileptogenicity Index (EI), 0.73 for the high-frequency energy ratio (HFER), 0.80 for ictal high-gamma power (HGPOW), and 0.69 for pre-ictal power (BASEPOW); all were above 0.5 (p < 1e-4). HGPOW exceeded BASEPOW by 0.11 (95% CI 0.04-0.19, p = 0.007), while EI and HFER did not. The HGPOW advantage did not replicate in seizure 2 (25 patients; difference +0.007, p = 0.92). Across 26 patients, ictal channel maps were less stable between seizures (median Spearman rho 0.28-0.42) than the pre-ictal power map (0.95). EI overlap with treated pairs did not distinguish Engel I from Engel II-IV outcomes. Because clinician SOZ labels came from the same ictal recordings, localization AUCs are partly circular. The study was single-center, included at most two seizures per patient, and had small outcome groups; the observed results do not establish outcome prediction or generalization beyond this dataset.

stereo-EEG · epilepsy · seizure-onset zone · epileptogenicity index · high-gamma power · signal localization · surgical outcome