In 4 resting SEEG cases, bipolar referencing reduced near-contact high-gamma correlation to 0.03

We compared spatial correlations in resting stereoelectroencephalography (SEEG) under monopolar, common average (CAR), and adjacent-contact bipolar references in four subjects from one centre. For each subject, we analyzed a 60-second rest window at 2048 Hz and measured zero-lag signal and envelope correlations across five frequency bands. In as-recorded monopolar high gamma (70–145 Hz), median absolute signal correlation was at least 0.92 even for contact pairs more than 50 mm apart. CAR removed 92–99% of high-gamma variance per contact. For pairs 5–10 mm apart, median high-gamma signal |r| was 0.95 monopolar, 0.25 CAR, and 0.03 bipolar; high-gamma participation ratio divided by channel count was 0.012, 0.45, and 0.74, respectively. With CAR, fitted correlation decay lengths were approximately 8 mm for theta to beta and 5 mm for high gamma. For near pairs, bipolar |r| was below CAR and CAR below monopolar in all four subjects for all bands and both measures, except high-gamma envelope correlation, where this ordering occurred in three of four. Four subjects allow no sign-test p-value below 0.0625; these counts describe consistency, not statistical significance. The common high-gamma signal may reflect the reference contact, muscle, or amplifier noise; this analysis does not identify its source. Results are limited to four subjects, one centre and amplifier, and 60 seconds of rest per subject. Zero-lag linear correlation does not measure decodable information, and task-evoked correlations may differ.

stereoelectroencephalography · SEEG · intracranial EEG · signal correlation · common average reference · bipolar reference · high gamma · brain-computer interface