Pre-click SEEG high-gamma activity detected change of mind but added no demonstrated value beyond kinematics in 15 patients

We tested whether 70–150 Hz SEEG high-gamma activity (HGA) detected a change of mind during mouse reaches after random-dot motion decisions, and whether HGA added information beyond four kinematic covariates. The analysis used archived author HGA window features for 15 patients from OpenNeuro ds006253 and followed a plan written and hashed before analysis. With five contiguous time-ordered folds, HGA in the 0.5 s before the click gave a mean held-out AUC of 0.616 (95% bootstrap CI 0.552–0.676); 12 of 15 patients had AUC above 0.5 (exact one-sided sign-flip p = 0.0021; Holm-adjusted p = 0.0063). HGA in the 0.5 s after movement onset did not detect a change of mind (mean AUC 0.523; adjusted p = 0.280). A behaviour-only model using movement duration, reaction time, peak velocity and motion coherence had mean AUC 0.867. Adding HGA changed mean AUC by +0.003 (95% CI −0.011 to 0.025; 4 of 15 patients improved; p = 0.477). Thus the pre-click HGA difference was not shown to add information beyond kinematics. Exploratory results indicate that the pre-click window usually occurred during movement and that change-of-mind reaches lasted about twice as long; this possible explanation was not tested. The analysis was offline and within patient, used three fixed 0.5 s windows, and did not validate dataset labels. An added value smaller than about 0.02 AUC cannot be excluded.

SEEG · high-gamma activity · change of mind · movement · brain-computer interface · cross-validation · AUC · OpenNeuro