Reach-duration matching reduced SEEG change-of-mind detection to a nonsignificant group effect in 14 patients

We tested whether stereo-electroencephalography (SEEG) activity in the 70–150 Hz band detected changes of mind (CoM) during reaches in a mouse motion-decision task. We used archived author-provided window features and behavior labels from 14 of 15 patients in OpenNeuro ds006253. Because CoM reaches were longer, we matched each CoM trial to two non-CoM trials by log reach duration, then evaluated a shrinkage linear discriminant decoder with five contiguous folds. In the primary test, matched activity from the 0.5 s before the response click had mean held-out AUC 0.559 (95% CI 0.489–0.629); 9/14 patients were in the predicted direction, and the exact one-sided group sign-flip test was not significant (p = 0.068). Movement-onset activity also did not detect CoM (mean AUC 0.486, p = 0.733). Reach duration alone had mean AUC 0.764 across all trials and 0.502 after matching. Matching reduced the pre-click neural AUC by 0.054 (p = 0.019; Holm-adjusted p = 0.058). These offline results do not support a general CoM signal from fixed pre-click windows across patients. Event-aligned windows may not be available to a real-time brain-computer interface, the CoM time within each reach was unknown, and the sample and matched trial counts were small.

stereo-electroencephalography · high-gamma activity · change of mind · brain-computer interface · reach duration · linear discriminant analysis · OpenNeuro ds006253