Pre-stimulus SEEG high-gamma activity did not decode the previous trial’s choice or outcome in 15 patients

In 15 patients performing random-dot motion decisions with a mouse, a shrinkage linear discriminant analysis of archived 70–150 Hz stereo-electroencephalography (SEEG) activity during the 0.5 s before each stimulus did not decode the previous trial’s choice or outcome. Across patients, previous-choice mean AUC was 0.505 (95% t CI 0.477–0.533; 10/15 above 0.5; exact one-sided sign-flip p = 0.36; Holm-adjusted p = 0.44); previous-outcome mean AUC was 0.513 (95% t CI 0.478–0.548; 9/15 above 0.5; p = 0.22; Holm-adjusted p = 0.44). Within current-choice strata, previous-choice mean AUC was 0.503 (95% t CI 0.476–0.531; p = 0.40). Mean choice-repeat rate was 0.52, and patient repeat rate did not predict previous-choice AUC (Spearman ρ = −0.05, p = 0.85). The test is a null result: for this decoder and window, the confidence intervals exclude mean AUC above 0.534 for choice and 0.549 for outcome. It does not exclude small effects in single channels, effects in other bands or windows, or effects detectable with nonlinear decoders. The offline analysis used one session per patient, all bipolar pairs, five contiguous time-ordered folds, training-fold z-scoring and cubic trial-index detrending, and Ledoit–Wolf shrinkage LDA. For decision decoders on this task, the pre-stimulus window did not provide a usable history prior or previous-trial leakage above about AUC 0.55.

stereo-electroencephalography · SEEG · high-gamma activity · trial history · choice decoding · outcome decoding · random-dot motion · linear discriminant analysis · brain-computer interfaces