Movement-side decoding was weak on average in eight SEEG patients, with only one new patient tested

We repeated a left/right mouse-choice decoding analysis of 70–150 Hz SEEG activity from the MetaRDK random-dot motion task. The original seven patients were re-analysed with unchanged extraction and analysis code, and their AUC and permutation values matched the earlier report to the last digit. A pre-registered extension planned up to 15 new patients, but low network speed allowed only sub-14 to finish. The primary new-patient test therefore could not give a result: with one patient, the smallest possible one-sided sign-flip p value is 0.5. In sub-14, movement-window activity (0–0.5 s after movement onset) decoded the chosen side on correct trials (AUC 0.783, permutation p = 0.005; 164 trials, 77 bipolar channels); on 36 error trials, the decoder followed the movement (AUC 0.857). In the secondary pooled analysis of eight patients, mean movement-window choice AUC was 0.569 (6/8 above 0.5; exact one-sided sign-flip p = 0.031), without correction for five windows and including the seven patients that motivated the question. A positive control distinguished movement from pre-stimulus rest (mean AUC 0.806; 8/8 above 0.5), with higher movement-versus-rest than choice AUC in every patient. Movement time was easy to decode in this task, while movement side was strong in only a few patients (sub-14 and sub-04). This small extension is not a replication: electrode coverage may affect side decoding, and many more patients are needed to estimate the group effect. Side AUC may reflect motor or somatosensory activity of the moving hand, or eye movements; these sources were not separated.

SEEG · high-gamma activity · left-right choice · brain-computer interfaces · linear discriminant analysis · AUC · random-dot motion · replication