In 36 SEEG patients, bipolar zero-lag correlation was near zero beyond 15 mm, while recorded-reference correlation remained above 0.1 across distances

We measured how zero-lag correlation between SEEG channels changes with distance. We used 36 patients from OpenNeuro ds004100 (snapshot 1.1.3). Each patient gave one interictal segment of 20-60 s. Contact positions were in fsaverage space. We used three montages: recorded reference (REF), common average reference (CAR) and adjacent-contact bipolar (BIP). We used six bands from 1 to 115 Hz and the 65-115 Hz amplitude envelope. A plan with four tests was written before analysis. With REF, r was more than 0.1 at all distances. The median r beyond 40 mm was 0.23 to 0.50, and it was highest at 65-115 Hz. With CAR, r fell below 0.1 at about 35 mm and was slightly negative far away. With BIP, the median r at 5-10 mm was 0.10 for 1-4 Hz and 0.01 for 65-115 Hz, and r was near 0 at 15 mm or more. The positive control was supported: for 1-4 Hz, far r was 0.34 higher with REF than with BIP (36 of 36 patients, Holm p = 0.0002). BIP r at 5-10 mm was 0.09 lower for 65-115 Hz than for 1-4 Hz (Holm p = 0.0002). The test of a shorter BIP decay length at 65-115 Hz was not supported (p = 0.19). The exponential fit did not work for BIP: the decay length was at the 1-mm grid limit in up to 22 of 36 patients. Thus we do not interpret the decay-length test for the envelope. Removal of seizure-onset channels did not change the results. Limits: template positions, short segments of unknown state, and epilepsy patients only. The package has all code, per-patient values, byte ranges with SHA-256 hashes, and run instructions.

SEEG · intracranial EEG · zero-lag correlation · electrode montage · bipolar reference · common average reference · OpenNeuro