Post-movement SEEG high-gamma activity decoded choice in three new patients, but pre-movement decoding was not confirmed

We tested whether 70–150 Hz SEEG high-gamma activity decodes left/right choice in three patients not used to develop the method. Using the original analysis code unchanged, a Ledoit–Wolf shrinkage LDA on log mean HGA features with five time-ordered cross-validation blocks, and a pre-registered test with 1,000 label permutations, the mean AUC in the 0–0.5 s after movement onset was 0.610 (null 95th percentile 0.569, p = 0.005, Holm-adjusted p = 0.010). The pre-movement mean AUC was 0.544 (null 95th percentile 0.564, p = 0.116, Holm-adjusted p = 0.116) and was not confirmed. Two new patients showed above-chance post-movement decoding (sub-03 AUC 0.622; sub-14 AUC 0.783), while sub-19 did not (AUC 0.424). Across all ten patients, the post-movement mean AUC was 0.560 (seven above 0.5; exact sign-flip p = 0.039); on error trials, a decoder trained on correct trials predicted chosen side (mean AUC 0.602, two-sided p = 0.010). Because choice and movement direction coincide in this task, the result does not distinguish decision coding from motor coding. The primary permutation test is valid for these fixed patients and does not generalize to the population. Only the author-provided 70–150 Hz derivative was used; the HGA pipeline was not checked.

SEEG · high-gamma activity · choice decoding · movement · mouse motion decision task · shrinkage LDA · replication · brain-computer interface