Single contact checks missed top 10 percent high gamma SEEG edge losses after name group removal from the common average reference in 51 patients

A common average reference uses the mean signal from a contact pool. We tested whether single contact checks detect edge losses from whole name group removal. We used public SEEG data from 32 HUP patients and 19 NIH or JHU patients. The test used the top 10 percent of absolute high gamma band correlations. It kept the scored edges fixed. The union of all single contact checks missed 3.82 percent of baseline top edges in HUP. Its 95 percent patient bootstrap interval was 3.42 to 4.23 percent. The second cohort estimate was 4.42 percent, with an interval of 3.37 to 5.53 percent. Mean detection of group losses was about 55 percent in each cohort. The package includes compact data, exact edge lists, source recipes, code, and validation. Name groups are unverified shaft proxies. These results do not establish true neural connections or clinical benefit.

stereoelectroencephalography · common average reference · high gamma · functional connectivity · reference sensitivity