HUP SEEG replicates patient-level similarity in relative power and exponent, while broadband power shows a smaller effect

An analysis of interictal SEEG features from 37 patients and 3,756 bipolar pairs in the HUP cohort tested whether pairs from the same patient are more similar than pairs from different patients at matched fsaverage distances. At 40 mm or more, the estimated fraction of across-patient squared difference absent within a patient (rho) was 0.44 for relative theta power (95% CI 0.32–0.53), 0.41 for relative alpha power (0.33–0.48), 0.46 for relative beta power (0.39–0.53), and 0.28 for the aperiodic exponent (0.20–0.35); all four had Holm-adjusted p=0.0005. These estimates were near those in the RAM cohort. The replication criterion was met for all five features. Broadband log power had a smaller patient-level component in HUP (rho=0.11, 0.06–0.17; Holm p=0.0011) than reported in RAM (0.52). The relative-power and exponent results persisted in opposite-hemisphere pairs, after removing seizure-onset and resected pairs, and in a second interictal segment. The findings support treating the patient as the sampling unit, or including a patient term in mixed models, for spectral-shape analyses. The study included one center and 37 patients; fsaverage registration error can inflate similarity estimates at short distances, and the cause of the patient-level component is unknown.

stereo-EEG · SEEG · intracranial EEG · spectral power · aperiodic exponent · patient-level effects · semivariance · replication · HUP · OpenNeuro