A theta peak rule showed no clear gain in 50 ms sine phase forecasts against offline labels for two contacts per patient in 36 SEEG patients

A theta peak contact rule showed no clear mean gain in this fixed test. The study used public OpenNeuro ds004100 version 1.1.3 data from 36 patients. Each patient had two good SEEG candidates, selected by contact name. Run 01 supplied the spectral score and frequency. A causal sine model forecast the run 02 offline theta phase label about 50 ms ahead. Mean selected contact error was 82.74 degrees. Gain over equal contact choice was 0.64 degrees. The 95 percent patient bootstrap interval was -0.55 to 1.85 degrees. The theta power comparison and three fixed sensitivity tests also gave intervals across zero. Mean gain in ten spectrum controls per patient was 0.71 degrees. The result does not prove a zero effect. It applies to this predictor, these two candidates, and this offline label. Recording order between files is not verified. The package preserves complete native contact traces, scored forecasts and labels, source versions and hashes, code, and reproduction instructions. This is a new endpoint in a reused cohort, not an independent cohort replication.

SEEG · theta phase · phase forecasting · contact selection · intracranial EEG · negative result · OpenNeuro