SEEG high gamma did not reliably decode changes of mind in four patients

We tested whether 70–150 Hz high gamma activity (HGA) in stereo-electroencephalography (SEEG) could predict or detect changes in mouse movement direction during random-dot motion decisions. In four patients from OpenNeuro ds006253 with at least 15 change-of-mind trials, we used within-patient L2 logistic regression, five contiguous time-ordered folds, and 500 label permutations. Median out-of-fold AUC was 0.515 before movement onset (−0.5 to 0 s; three of four patients above 0.5; group p = 0.44, Holm p = 0.44) and 0.538 during the first second after movement onset (three of four above 0.5; group p = 0.19, Holm p = 0.38). No patient had permutation p < 0.05; the highest was 0.628 (sub-23, post-onset; p = 0.056). HGA did not improve on a behavior model of coherence, movement latency, and correctness: median AUC gain was −0.007 before and +0.004 after movement onset. Exploratory no-fit movement-latency AUC was 0.609–0.648 in the four patients, with slower starts accompanying changes of mind. These results do not establish that the brain lacks a change-of-mind signal. They are limited by four patients, sparse trials, dataset-defined change labels, incomplete epochs, different event timing, possible motor signals in the post-onset window, and uncorrected exploratory analyses. The smallest possible one-sided exact group p-value with four patients was 0.0625.

stereo-electroencephalography · high gamma activity · change of mind · brain-computer interface · mouse movement · motion decision · cross-validation · negative result