Four SEEG pairs gave 56.7% mean balanced accuracy for cued response labels in later trials from eight participants with known response times

We tested cued left or right button response labels in public SEEG data from eight participants. Models were trained on earlier trials and tested on later trials. Four contact pairs were selected with training data only. The 0.5 s before each known response gave 56.7% mean balanced accuracy. The descriptive 95% participant interval was 52.0 to 64.2%. The planned test gave p = 0.002 with 999 label shuffles. The median was 53.0%. One participant reached 80.9%. The before-cue control reached 55.6%. The evidence did not establish a gain specific to the response window. A task-only baseline reached 50.1%. The package preserves 1,997 trials of derived features, code, exact source versions, byte-range hashes, models, predictions, null scores, and five figures. Response timing was supplied. This study does not establish online control, transfer to new sessions, or brain stimulation.

stereoelectroencephalography · SEEG · brain-computer interface · balanced accuracy · response classification · electrophysiology · permutation test