Phase randomization after a bipolar montage rejects shared-contact controls in two public SEEG excerpts

This study tests the order of phase randomization and bipolar referencing in stereoelectroencephalography (SEEG) correlation tests. It uses two 60-second excerpts from OpenNeuro ds005670, version 1.0.0. Controls retain the observed contact spectra and have independent contact phases. At a 5 percent per-edge threshold, tests that randomize phases after the montage rejected 99.70 and 99.52 percent of shared-contact controls in participants 01 and 02. Tests that randomize phases before the montage rejected 5.10 and 5.23 percent. Both models gave rejection rates near 5 percent for disjoint control classes. The covariance identity for bipolar channels explains why independent contact phases retain covariance from a shared physical contact. A rejection under the after-montage null therefore does not establish contact coupling. The package provides prepared signals, source retrieval instructions, code, all candidate-edge tests, and numerical checks. The result applies to this conditional Fourier phase model and these two excerpts. It does not identify true brain connections or estimate a population error rate. The recordings have existing references, and their phases do not reconstruct unreferenced neural voltage.

stereoelectroencephalography · SEEG · bipolar montage · phase randomization · surrogate data · shared-contact control · Pearson correlation · calibration · intracranial EEG · Fourier phase