Adding alpha and beta power to high-gamma power at the same eight pairs did not show a clear mean error reduction in later trials from eight SEEG recordings

We tested whether alpha and beta power improve a reader that uses high-gamma power. We used 2014 trials from eight public SEEG recordings in OpenNeuro ds004473, version 1.0.2. The primary comparison used the same eight bipolar pairs within each person. Earlier trials set the pairs, scaling, and model settings. Later trials tested each reader. Mean Brier probability error was 0.17964 for all three bands and 0.19288 for high gamma alone. The paired mean change was -0.01325. Its descriptive 95 percent person-bootstrap interval was -0.03041 to +0.00255. Four people improved and four worsened. Thus, the test did not show a clear mean reduction. It did not show equal performance or a lack of useful signal in the added bands. The package provides source features, raw audit ranges, model code, saved readers, predictions, figures, source hashes, and repeat checks. The labels describe offline task states in reused clinical records. The test does not measure continuous command detection or brain stimulation.

stereoelectroencephalography · brain-computer interface · Brier score · high-gamma power · alpha power · beta power · feature selection · logistic regression · temporal validation · OpenNeuro ds004473 · probability calibration · neural decoding