Matching removed neighbor counts reduces the mean excess in beta SEEG channel choice changes in 21 patients

We tested a graph control for SEEG channel choice. We used 35 public records and 70 event graphs from 21 patients in OpenNeuro ds006519 version 1.0.0. Channel score was mean absolute beta correlation. Removing edges between bipolar channels with a common contact changed the highest channel at a patient mean rate of 29.3 percent. The excess over random removal was 14.1 percentage points when only the edge count was fixed. It was 7.0 points when each channel lost the same number of neighbors as in the observed mask. The 95 percent patient bootstrap interval was -2.8 to 16.6 points. The primary sign-flip test gave p = 0.184 under independent patients and symmetric null differences. Keeping removals within each shaft gave an excess of -5.7 points. Thus the graph constraints changed the interpretation of channel sensitivity. We preserve a proved uniform sampler, 420,000 masks, scores, signal arrays, source hashes, code, and three figures. A complete second run gave identical numeric output. This audit does not measure an improvement in a brain-computer interface.

stereoelectroencephalography · SEEG · intracranial EEG · beta correlation · graph control · channel selection · shared contacts · degree matching · electrode shafts · imaginary coherency