Under an assumed right-hand task, ten left-hemisphere SEEG channels did not outperform ten right-hemisphere channels for mouse-movement detection in 15 patients

We compared movement-versus-rest decoding from equal numbers of left- and right-hemisphere stereo-EEG (SEEG) channels in 15 patients performing a mouse-based random-dot decision task. The dataset and its associated paper do not report which hand patients used; we assumed right-hand use, so left and right correspond to contralateral and ipsilateral only under that assumption. Using the unchanged archived Ledoit-Wolf linear discriminant decoder and 10 random channels per hemisphere (mean across 200 subsets), held-out AUC was 0.678 for left and 0.666 for right (difference +0.012; 95% bootstrap CI -0.037 to +0.059; exact stratified permutation p = 0.65; 10 hemispheres per side). We found no left/right difference, and the confidence interval excludes a left advantage larger than 0.06 AUC. In five bilateral patients, the left-minus-right difference at equal channel count was +0.030 (3/5 positive; exact sign-flip p = 0.50). Both hemispheres together exceeded the left alone in all five (+0.121; p = 0.0625, the smallest possible two-sided value for n = 5). Removing the non-neural marker channel changed mean AUC from 0.833 to 0.832 (largest patient change 0.0098; p = 0.19); marker alone gave mean AUC 0.55. Limits include the assumed hand, unequal sampled regions, five bilateral patients, one offline session per patient, one time-ordered train/test split, and no closed-loop test. These results are consistent with a prior key-press study in another dataset.

stereo-EEG · brain-computer interface · movement decoding · hemisphere · high-gamma activity · mouse movement · Ledoit-Wolf LDA · OpenNeuro