The 95% interval for SOZ AUC gain from patient ranks included zero for a fixed logistic model in 18 test SEEG patients

We tested whether patient ranks improve seizure onset zone (SOZ) contact ranking in public SEEG data. The fixed cohort had 4,252 contacts from 36 patients in OpenNeuro ds004100 version 1.1.3. We fitted two fixed logistic models on 18 patients and tested the other 18. Mean patient AUC was 0.7373 for raw features and 0.7481 for patient ranks. The paired gain was 0.0108. Its 95% patient bootstrap interval was -0.0131 to 0.0370. The exact sign-flip p value was 0.4292. The interval conditions on the fixed training group and assumes independent test patients. The sign-flip test also assumes symmetric differences under its null. The study did not show a clear gain. Repeated patient folds gave a mean gain of 0.0019. The package preserves code, inputs, predictions, source versions, hashes, checks, and reproduction instructions. Clinical labels, inherited contact selection, short samples, and one epilepsy cohort limit its scope.

stereoelectroencephalography · seizure onset zone · patient rank scaling · contact ranking · AUC · patient-separated testing · spectral features