Adding an insula high-gamma term after movement increased mean confidence prediction error in a fixed model across trial blocks in 14 SEEG patients

We tested whether mean insula high-gamma activity in the 400 ms after first mouse movement improved confidence prediction. The baseline used behavior and mean insula activity in the 400 ms before movement. We used public MetaRDK SEEG derivatives from OpenNeuro ds006253, release 1.0.3. A fixed anatomy rule selected 14 patients, 2,971 trials, and 78 bipolar pairs. Five separate trial blocks tested each patient model. The mean patient increase in mean squared error was 0.51%. The 95% patient bootstrap interval for error reduction was -1.45% to +0.33%; the exact sign-flip p value was 0.3081. The result does not establish a benefit or the absence of confidence information. A fixed forward-time check reduced error, but both models had poor absolute skill. The package preserves code, compact data, channel maps, predictions, model parameters, inference draws, source versions, public source checks, and reproduction instructions. Source smoothing and prior normalization limit timing and raw-data pipeline claims. This is a retrospective test in the same clinical cohort as the source work.

stereoelectroencephalography · confidence prediction · insula · high-gamma activity · retrospective prediction · clinical neuroscience