Reference agreement reduced mean alpha peak recurrence probability error by 1.8% in fixed models for paired 60-second SEEG excerpts from 24 test patients

We tested whether agreement between two references helps to predict alpha peak recurrence in another supplied SEEG file. We used OpenNeuro ds004100 version 1.1.3. We fit models with 12 patients and tested them on 24 other patients. Reference agreement reduced mean patient Brier score from 0.227037 to 0.222958, a relative reduction of 1.80 percent. The absolute gain was 0.004079. Its conditional 95 percent patient bootstrap interval was 0.000402 to 0.007843. Fifteen patients had lower error; nine had higher error. Mean AUC fell slightly. The biological interval between paired 60-second excerpts is unknown. The result concerns these fixed models and matched contact selection, which used both files. The package preserves 36 paired spectrum files, public byte retrieval recipes, code, predictions, and validation records. All 72 raw excerpts were retrieved and rebuilt. Final numerical results matched.

SEEG · alpha peak · reference agreement · recurrence · Brier score · patient-separated prediction · spectral analysis · epilepsy