High-gamma choice decoding was weak and not significant in five SEEG patients

We tested whether a left/right decoder trained on correct trials from stereo-electroencephalography (SEEG) high-gamma activity follows a patient's reported choice or the motion stimulus on error trials. We analyzed five patients from OpenNeuro ds006253 using binned 50 ms, 70–150 Hz high-gamma epochs and behavior tables archived in SCI contribution 8d5572e9. In the preregistered primary window, 0–0.5 s after movement onset, mean error-trial AUC against choice was 0.567; four of five patients had AUC above 0.5, but the exact one-sided sign-flip test was not significant (p = 0.063). No window showed reliable error-trial evidence for stimulus coding (p ≥ 0.59). Correct-trial decoding was weak (mean AUC 0.517–0.552); only the stimulus-locked window was above chance in all five patients (p = 0.031, the smallest possible group p). These results provide no significant support for the primary choice-coding hypothesis and no evidence of stimulus coding. The signal was too weak for a useful single-trial choice readout in this sample. The analysis was offline, used clinical electrode coverage, and included five patients; error trials may differ in difficulty and attention. Cross-validation repeats and correct-trial permutations were reduced from the preregistered plan because of a compute limit.

stereo-electroencephalography · high-gamma activity · choice decoding · motion decision · brain-computer interface · error trials · linear discriminant analysis