One fixed cross-patient threshold retained most SOZ separation in 43 NREM SEEG patients

We asked whether a single threshold across patients preserves seizure-onset-zone (SOZ) separation by three interictal non-rapid-eye-movement (NREM) stereo-electroencephalography markers, or whether within-patient normalization is needed. Using archived per-pair features for 43 patients (728 bipolar pairs, 160 SOZ pairs; 90 s NREM), pooled raw-value AUCs were 0.659 for log 80–300 Hz power (HFA), 0.678 for log slow-wave-phase to HFA modulation index (MI), and 0.672 for log 0.5–4 Hz delta power. Within-patient z-scores yielded AUCs of 0.674, 0.692, and 0.645. Differences were small and not significant; both preregistered hypotheses were unsupported. Raw MI varied more between patients than raw HFA power (non-SOZ ICC 0.37 vs 0.18). Leave-one-patient-out thresholds had sensitivity 0.68–0.71 and specificity 0.49–0.64, so these are not clinical-grade rules. Clinical SOZ labels are not independent of interictal activity; only 90 s per patient and one dataset family were analyzed. Confidence intervals do not exclude normalization gains of about 0.05–0.09 AUC. The analysis used archived features, so feature-extraction errors carry over.

stereo-electroencephalography · seizure onset zone · interictal · NREM · high-frequency activity · modulation index · AUC · normalization · epilepsy