SEEG response detection across two keys of one hand reached 69.7% mean balanced accuracy in event-aligned later trials from eight patients with epilepsy

We tested whether SEEG response detection transferred between two keys of one hand. The data came from eight patients with epilepsy in OpenNeuro ds004473 version 1.0.2. Each detector used eight bipolar pairs. Early trials from one recorded key supplied training. Later trials from the other key supplied the test. Response and rest windows used known event times. Training counts and fast/slow conditions were matched. Mean cross-key balanced accuracy was 69.7%. Its 95% patient bootstrap interval was 63.6% to 76.0%. The exact one-sided patient sign-flip value against chance was 0.003906. Mean same-key accuracy was 69.4%. The paired cross-key change was +0.28 percentage points, with an interval of -2.15 to +2.58. The interval does not establish equal performance. The result supports an offline calibration test across the two recorded keys. Shared task states can contribute. It does not establish cue-independent motor intent or online control. The package preserves complete inputs, all predictions, code, figures, hashes, and raw-range checks.

stereoelectroencephalography · brain-computer interface · response detection · calibration transfer · balanced accuracy · epilepsy