Adding high gamma gave a mean key side loss gain of 0.0016 nats in a fixed test of later SEEG trials from eight patients

This study tested whether high gamma power improves left versus right key prediction beyond beta power. It used 2,030 trials from eight patients in public OpenNeuro dataset ds004473, version 1.0.1. Within each patient, the first half trained a fixed logistic model and the later half tested it. The mean loss gain from adding 70–110 Hz power to 13–30 Hz power was 0.001552 nats per trial. The descriptive 95% patient bootstrap interval was -0.012775 to 0.016954 nats. The exact two-sided patient sign-flip test gave p = 0.828125. This result does not establish a clear mean gain for this fixed model. A paired window control reproduced the earlier gain of 0.082666 nats per pair. The penalty and pre-cue subtraction affected the side gain. This is a secondary analysis of a known cohort. Cue direction usually agrees with key side, and response windows include time after the key. The package preserves the local plan, features, metadata, code, model weights, predictions, statistics, figures, and retrieval recipes. A fresh rebuild matched all feature arrays and 4,076 raw byte-range hashes. A separate spectral check and a repeated offline analysis passed. Full EDF file hashes were not checked. The study does not establish population equivalence or an online motor decoder.

stereoelectroencephalography · key-side prediction · high gamma · beta power · logistic regression · fixed forward split · brain-computer interfaces