Adding beta and high-frequency log power products did not show a clear Brier gain after a task change in eight people with eight SEEG pairs each

We tested whether beta and high-frequency log power products improved a stereoelectroencephalography (SEEG) reader of pre-response and rest windows after a task condition change. We used public OpenNeuro ds004473 version 1.0.2 data from eight people with epilepsy. All readers used eight pairs, three power bands, and 33–70 early calibration trials per person. Pair choice and penalty choice used training data. Mean Brier score was 0.2154 with separate squared band effects and 0.2156 with same-pair products. The mean gain was -0.0002; the descriptive 95% person bootstrap interval was -0.0019 to 0.0019. The exact sign reference was p=0.867, under independent symmetric person differences. The test did not establish a gain. In the planned earlier-rest test, products increased Brier error in all eight people. The Holm-adjusted sign reference was p=0.023. The package includes features, source versions and hashes, raw audit ranges, fitted states, predictions, figures, and replay code. This is an exploratory offline test in a reused cohort. It does not test live control, phase-amplitude coupling, or stimulation.

stereoelectroencephalography · SEEG · brain-computer interface · log power · Brier score · task transfer · logistic regression · exploratory study