In eight SEEG subjects, beta and high-gamma power maps differed and beta-selected channels were not shown non-inferior

We tested whether beta (13–30 Hz) decreases and high-gamma (70–150 Hz) increases identify the same bipolar SEEG channels during a cued key press, and whether beta can select channels for decoding. Using archived features for eight subjects from OpenNeuro ds004473 v1.0.2, each map had split-half reliability of 0.34–0.93, but the disattenuated map correlation was near zero (mean rho = −0.026; rho < 1 in all eight subjects; Holm p = 0.012). The across-channel correlation was not negative (mean r = +0.034; p = 0.60). With channel selection on trials 1–100 and testing on trials 101+, eight beta-selected channels yielded mean HG+beta LDA AUC 0.798 versus 0.820 for HG-selected channels; non-inferiority at a 0.02 AUC margin was not shown (mean difference −0.022; p = 0.56). Beta-only decoding on beta-selected channels had mean AUC 0.771. In this task and these windows, the maps largely occupied different channels, and beta maps did not replace HG maps for channel selection. The analysis is offline and limited to eight subjects, one session each, fixed cued windows and fixed frequency bands; the rest window precedes the fixation cross, so the beta contrast may include cue and attention effects. The analysis used previously archived features and did not re-extract them from the original recordings.

stereo-EEG · beta power · high-gamma · brain-computer interface · channel selection · linear discriminant analysis · motor task