Common average reference changed a mean of 15.4 percent of top 10 percent zero-covariance control hubs in 1–80 Hz bipolar SEEG samples from 36 patients

We tested whether a contact covariance control keeps the same bipolar SEEG hubs after a reference change. We used paired 60-second interictal samples from 36 patients in OpenNeuro ds004100, version 1.1.3. The control removes all covariance between contacts after the selected reference. Common average reference changed a mean of 15.4 percent of its top 10 percent hubs at 1 to 80 Hz. The descriptive 95 percent patient bootstrap interval was 12.1 to 18.7 percent. Measured bipolar signals and hub lists stayed the same. An exact proof explains why diagonal removal can break reference cancellation. A fixed control retains cancellation when its full covariance is carried through the reference transform. We verified public metadata and rebuilt all 72 contact covariance files from pinned EDF ranges. The package preserves inputs, source code, checks, figures, and reproduction instructions. This same-cohort method audit does not estimate a neural artifact fraction or clinical performance.

stereo-EEG · bipolar montage · common average reference · covariance control · network hubs · reference sensitivity · connectivity · reproducibility