A fixed low frequency score ranked seizure onset contacts less well at named contact boundaries in 34 SEEG patients

We tested whether a fixed low frequency score ranked clinical seizure onset contacts above their named neighbors. We used two 120-second interictal SEEG records per patient from OpenNeuro ds004100, version 1.1.3. The score was log10 of median 2–4 Hz density divided by median 0.5–1 Hz density. The primary test used stored signals and 126 opposite-label pairs from 34 patients. Each patient had equal weight. The mean local ranking rate was 0.508. The mean AUC for all contacts in the same patients was 0.648. The paired local minus all-contact change was -0.140, with a 95 percent patient bootstrap interval of -0.217 to -0.063. The two-sided Holm p value was 0.00258. The local interval was 0.418 to 0.596. This interval does not prove equal performance to chance. Numbered neighbors were not verified tissue boundaries. A related paper that appeared during this study reported the same local mean from the same inputs. The two implementations agreed exactly. This is not evidence from an independent cohort. The present study adds the paired test, a contact-number baseline, gap tables, and an exact AUC decomposition. All fixed sensitivity results are reported, including stronger ranking under some choices. The package preserves source code, feature arrays, source versions, hashes, tables, figures, and retrieval instructions. A complete repeat gave identical numerical outputs. An independent raw check agreed for one recording. The evidence concerns short records and clinical contact labels. It does not establish treatment benefit.

SEEG · epilepsy · seizure onset zone · local ranking · low frequency power · contact-level AUC