SOZ detectability across patients is feature-specific beyond pair-level collinearity in awake SEEG cohorts

A pre-registered analysis tested whether per-patient detectability of the clinical seizure-onset zone (SOZ) is a shared patient trait across interictal features, beyond correlation expected from pair-level feature collinearity. In 59 awake patients from OpenNeuro ds004100 and ds003876, the correlation between per-patient SOZ AUCs for 80–150 Hz high-frequency activity (HFA) and spike rate was rho = -0.167, compared with a circular label-shift null mean of +0.069 (SD 0.132; one-sided p = 0.964). The two cohorts had the same direction. None of 15 awake feature pairs exceeded the null; 13 excesses were negative (exploratory). In sleep cohorts, observed correlations also did not exceed their null means: ds007118–20 HFA versus slow-wave-HFA coupling, rho = 0.571 versus 0.588 (n = 40), and ds005398 tonic HFA versus bursts, rho = -0.259 versus +0.306 (n = 18). These results are consistent with patient-to-patient AUC variation being mostly feature-specific or noise, rather than one shared patient property. Interpretation is limited by clinical labels, short clips, an assumed correspondence between file and shaft order in the shift null, noisy AUCs, and an approximate upper limit; the analysis does not identify causes of AUC variation.

SEEG · seizure-onset zone · interictal · high-frequency activity · spike rate · AUC · collinearity · circular shift null