SEEG waveform repeatability gave AUC 0.492 for agreed versus disputed responses in the UMCU21 record

This study tested whether waveform repeatability identified expert agreement about an evoked response. It used OpenNeuro ds004774 version 1.0.0, record UMCU21. The source selection contained 330 pulses, 33 surface stimulation sites, and six SEEG readout contacts. Two experts had finite labels for 188 cells. The fixed score compared the first five and last five pulse means. It subtracted the before-pulse correlation from the 10 to 90 ms after-pulse correlation. AUC was 0.492 for 38 agreed versus 38 disputed response cells. The descriptive 95% site bounds were 0.363 to 0.632. High repeatability occurred in 31 disputed cells. Response presence agreement was 79.8%, with kappa 0.529. One shared binary prediction, scored equally against both experts, had an exact pooled accuracy limit of 89.9%. The package preserves voltage, source labels, code, figures, all bootstrap draws, and reproduction instructions. Fresh raw checks and an exact second analysis passed. The result applies to one record. It does not establish equivalent groups, neural origin, or brain interface performance.

SEEG · evoked responses · waveform repeatability · expert agreement · AUC · intracranial electrophysiology · OpenNeuro