In 5 ds006253 patients, SEEG high-gamma 0-0.5 s after the response click did not separate error from correct decisions better than behaviour (no feedback, offline LDA; underpowered pilot)

We tested if SEEG 70-150 Hz activity after the response click separates error decisions from correct decisions in a random-dot task with no feedback (OpenNeuro ds006253, author high-gamma derivative). The plan was written and hashed before analysis. Only 5 of 7 planned patients finished downloading (02, 04, 09, 13, 20). Decoder: shrinkage LDA on log high-gamma per bipolar channel, 5 time-ordered folds, AUC per patient, exact sign-flip test over patients. Primary window (click 0-0.5 s): mean AUC 0.508, 3 of 5 above 0.5, p = 0.344; no patient was significant alone. With behaviour removed by regression: mean AUC 0.503. A decoder on 4 behaviour values (coherence, RT, mouse speed, movement duration) gave mean AUC 0.574, higher than all neural windows. Before the click (-0.5-0 s), mean AUC 0.554, 5 of 5 above 0.5, uncorrected p = 0.031, Holm p = 0.156. Deviation: the confidence scale appeared about 0.22 s after the click, not 1 s as planned. Limits: pilot, clinical coverage, high-gamma only.

SEEG · error decoding · high-gamma · brain-computer interface · random-dot motion · LDA · confidence · OpenNeuro ds006253 · pilot study