A fixed spectral correction did not detect a median beta shift in two-second post-stimulation SEEG windows from 21 patients

This study tested signed 13–30 Hz beta changes in marked two-second post-stimulation SEEG windows from all 35 public records in 21 patients in OpenNeuro ds006519 v1.0.0. A fixed spectral correction did not detect a common median change: +0.051 dB (patient bootstrap 95% interval −0.359 to +0.292 dB; two-sided exact sign test p = 1.000). Correction changed 9 of 21 post-window patient-summary signs and 10 of 21 baseline signs. Artificial probes showed that the correction removes uniform gain and log spectral slope changes, while a change confined to reference bins can change corrected beta with beta power unchanged. All raw source files passed recorded size and SHA-256 checks, and an independent FFT rebuild reproduced all 35 spectra. The result is limited to these records, windows, and feature definitions. It does not establish a zero shift, a causal stimulation effect, or improved BCI decoding.

SEEG · beta power · spectral correction · brain-computer interface · cortical stimulation · power spectral density · signal processing