In 36 HUP SEEG patients, the 30-45 Hz interictal aperiodic exponent did not identify the SOZ; absolute bipolar band power and transient rate did (120 s clips, no anatomical correction)

Question: which simple interictal SEEG features identify the clinical seizure onset zone (SOZ) in each patient? We used OpenNeuro ds004100 (HUP iEEG). We took the first 120 s of interictal run-01 from all 37 SEEG patients and made bipolar pairs of adjacent contacts. For each pair we calculated 20 spectral and time-domain features. For each patient we calculated the AUC for SOZ versus non-SOZ pairs. We tested the group AUC against 0.5 with a sign-flip permutation Wilcoxon test and a Bonferroni correction. The primary feature, the 30-45 Hz aperiodic exponent, did not separate SOZ pairs (median AUC 0.49, 18/36 patients above 0.5, p = 0.51). Absolute log power in all bands was higher in SOZ pairs. Delta power had a median AUC of 0.74 (31/36, corrected p = 4e-4) and total power had 0.72. A transient-rate detector gave 0.68 and line length gave 0.68. Relative ripple-band power was lower in SOZ pairs (AUC 0.32). The SOZ AUCs were not different between seizure-free (n = 20) and not-seizure-free (n = 16) patients (all p >= 0.12). Limits: short clips, no sleep control, and no grey/white-matter or regional normalisation. Thus the absolute-power effects can be partly tissue or location effects. The package has the code (numpy only), the hashes of the truncated EDF files, the channel-level features, and the per-patient AUCs.

SEEG · intracranial EEG · epilepsy · seizure onset zone · interictal · aperiodic exponent · band power · transient rate · AUC