Linear contact position explains part, but not all, of SOZ separation in 13 HUP SEEG patients

In 13 patients from one center, interictal bipolar SEEG features from ds005624 were compared with clinical seizure-onset-zone (SOZ) labels from ds004100. We tested whether linear contact-position covariates explain the separation in 1–150 Hz power and line length. Within each patient and segment, ordinary least squares predicted each feature from deeper-contact number, absolute x and y/z coordinates of the pair midpoint, and intercontact distance; we assessed the SOZ-versus-non-SOZ AUC of the residuals. The position-residual AUC was 0.644 (95% CI 0.547–0.727) for power and 0.634 (0.546–0.713) for line length (Holm p=0.014 for both). Position explained part of the separation: the AUC loss was 0.068 (0.028–0.113; p=0.007) for power and 0.038 (−0.003–0.080; p=0.11) for line length. Within-shaft AUCs were 0.697 (0.585–0.796) and 0.748 (0.662–0.821), respectively. Position alone separated SOZ pairs by contact number (AUC 0.678); in three patients with mostly deep SOZ contacts, residual AUC was near 0.5. Thus, these linear covariates did not fully explain SOZ separation in this sample, while contact position explained part of the power effect. Limits include 13 patients from one center, fsaverage rather than patient-specific coordinates, no gray/white matter labels, linear covariates, uncontrolled task state, and clinical SOZ labels that may not be independent of interictal activity.

stereo-EEG · seizure onset zone · contact position · interictal activity · AUC · epilepsy