An unchanged decoder predicted peak mouse velocity in 8 new ds006253 SEEG patients, but the effect was small and offline

In a preregistered out-of-sample replication, an unchanged ridge decoder using 70–150 Hz SEEG high-gamma activity from 0–0.5 s after mouse-movement onset predicted held-out log peak velocity in eight new OpenNeuro ds006253 patients. Mean held-out r was 0.146; six of eight patient values were positive, and the exact one-sided sign-flip test gave p = 0.039 (10/256 patterns; planned alpha 0.05). Four patients had individual permutation p < 0.05. The pre-stimulus control did not predict velocity (mean r = 0.023, p = 0.293). In the descriptive pooled sample of 15 patients, mean r was 0.132 (12/15 positive, p = 0.0034; patient-bootstrap 95% CI 0.059–0.204). The exploratory 200 ms peak latency did not replicate exactly. The effect was small (about 2% of trial-to-trial variance), measured offline in one session and task, and does not show that the signal is sufficient for BCI speed control. The decoder used author-derived HGA and mixed clinical SEEG contacts; activity after movement onset may include sensory feedback and motor-control signals.

SEEG · movement vigor · high-gamma activity · ridge regression · mouse movement · replication · brain-computer interface