Across two SEEG cohorts, HFA, line length, and 1–40 Hz power showed no excess SOZ-AUC spread; spike rate and 30–45 Hz exponent did in ds004100

We tested if per-patient SOZ AUCs of 10 interictal SEEG features vary between patients more than contiguous clinical labels give by chance. We used archived 120-s bipolar-pair features of 30-31 HUP patients (ds004100) and 29 patients from 4 centers (ds003876). The null for each patient was a circular shift of the SOZ labels in shaft order. The statistic was a Cochran-type Q with 20000 null draws. The plan was hashed before the run. We used Holm correction over 20 tests. For HFA (80-150 Hz), line length and 1-40 Hz power, the spread was not larger than the null (H 0.67-1.32, Holm p = 1 in both cohorts). Thus one fixed direction and one mean effect is sufficient for these features, and low-AUC patients agree with chance label placement. Spike rate varied about 2 times more than the null in ds004100 (H 2.02, Holm p 0.006), and in the same direction in ds003876 (H 1.69, Holm p 0.053). The 30-45 Hz aperiodic exponent had a patient-specific sign in ds004100 (H 2.24, Holm p 0.001; AUC < 0.5 in 48 % of patients), but not in ds003876. Limits: one awake clip per patient, clinical labels, an additive noise model on the AUC scale, and shifts that cross shaft boundaries. The exact dataset releases behind the archived feature tables were not recorded.

SEEG · epilepsy · seizure onset zone · interictal · AUC · patient heterogeneity · intracranial EEG · label shift · spike rate · high-frequency activity