Beta selection gave lower high-gamma reader accuracy than gamma selection in later trials from eight SEEG recordings

We tested whether beta suppression can select inputs for a high-gamma brain reader. We used eight public SEEG recordings from OpenNeuro ds004473 version 1.0.2. The analysis used 2,014 trials and 834 bipolar pairs. Early trials supplied maps and reader training. Later trials supplied the test. The mean rank correlation between early beta suppression and later high-gamma increase was -0.025. Its descriptive 95 percent patient interval was -0.177 to 0.149. The same-band correlations were 0.465 for beta and 0.448 for gamma. In a planned secondary test, eight pairs selected by gamma increase gave 67.3 percent mean balanced accuracy. Eight pairs selected by beta suppression gave 61.7 percent. Gamma selection gave higher accuracy in all eight recordings. The paired difference was 5.6 percentage points, with a descriptive 95 percent interval of 2.7 to 9.9 points. This reused clinical cohort supports direct gamma measurement for this input rule. It does not establish isolated motor intent or live BCI control. The package preserves figures, features, source records, all fitted readers, code, and repeat instructions.

stereoelectroencephalography · brain-computer interface · beta suppression · high gamma · input selection · neural decoding