Region control preserves SOZ separation by sleep HFA features in 17 patients; region explains 10–13%

In 17 UCLA SEEG patients with interictal sleep recordings, we tested whether gray-matter bipolar-pair separation of clinical seizure-onset-zone (SOZ) from non-SOZ pairs by 80–300 Hz log power (log_hfa) and delta-phase/HFA modulation index (mi) persisted after leave-one-patient-out regional normalization. The pre-registered analysis used 896 gray-matter pairs (159 SOZ) across 32 dataset atlas labels; 857 pairs (158 SOZ) had a regional norm. Residual SOZ/non-SOZ AUC was 0.684 (95% CI 0.598–0.763; 14/17 patients above 0.5; Holm p=0.003) for log_hfa and 0.626 (0.543–0.697; 15/17; Holm p=0.007) for mi. Region-only AUCs were 0.564 (p=0.32) and 0.528 (p=0.71). Regional normalization reduced AUC by 0.021 and 0.020, respectively; the estimated shares explained were 10% and 13%, with confidence intervals for the AUC differences including zero. Within-region AUCs were 0.681 (95% CI 0.532–0.812; p=0.030) and 0.613 (0.509–0.710; p=0.054). This differs from a prior awake HUP result in which region explained one half to two thirds of separation; cohort, state, features, and region definition differ, and this analysis did not identify which factor accounts for the difference. Limitations include dataset atlas labels without MRI checks, small region norm samples, clinical SOZ labels, 90 seconds of sleep per patient, features inherited from a prior analysis rather than recomputed here, 17 analyzed patients, wide confidence intervals, and insufficient mesial temporal data for an informative subgroup test. A small regional effect (D below about 0.06) is not excluded.

intracranial EEG · seizure-onset zone · high-frequency activity · phase-amplitude coupling · brain region · epilepsy · AUC · regional normalization