With a 2 s offline reference, new calibration reduced mean 50 ms theta phase error in four of seven people with SEEG rest records

We tested whether early post calibration improves a fixed pre reader in matching SEEG rest records. We used OpenNeuro ds006065 version 1.0.0, seven people, and 28 fixed electrode pairs. Each reader used only the past 0.5 s of voltage to forecast theta phase 50 ms ahead. A centered 2 s offline filter supplied the reference phase. On separate later post data, mean absolute error was 57.10 degrees for the fixed reader and 56.67 degrees for the new reader. The mean gain was 0.43 degrees. A person bootstrap gave a descriptive 95% interval from -1.74 to 2.48 degrees. The exact two sided sign flip probability was 0.625, under independent people and symmetric null signs. New calibration helped four people and increased error in three. The mean gain stayed below 0.5 degrees with 1 s and 3 s reference filters. These data do not establish a consistent benefit from this short calibration. They also do not establish that the readers are equivalent. The package supplies compact voltage data, fixed choices, pinned source URLs, hashes, models, scores, figures and reproducible code. This is an offline test of rest records. It makes no causal stimulation, behavior, clinical or hardware claim.

SEEG · theta phase · phase forecasting · calibration transfer · ridge regression · brain computer interface