Bipolar SEEG high-gamma envelopes were nearly uncorrelated at 5–10 mm, but fitted space constants did not separate high gamma from theta

We measured how correlation between stereoelectroencephalography (SEEG) signals changes with Euclidean distance between contacts in 8 subjects from OpenNeuro ds004473. For each subject, we analyzed 120 s from a task recording using monopolar signals as recorded, common-average reference (CAR), and bipolar pairs of adjacent contacts. We measured correlations of 100 ms mean log Hilbert envelopes and band-passed voltage in theta (4–8 Hz) and high gamma (HG, 70–150 Hz). For HG envelopes at 5–10 mm, median correlation was 0.04 for bipolar pairs and 0.43 for monopolar signals; bipolar values were lower in all 8 subjects (mean difference 0.39, exact one-sided sign-flip p = 0.0039). Normalized per-shaft participation ratio for HG envelopes was 0.89 for bipolar and 0.46 for monopolar signals; it was higher for bipolar in all 8 subjects (mean difference 0.37, p = 0.0039; CAR median 0.67). However, the primary hypothesis that bipolar HG envelope space constants were shorter than theta space constants was not supported: the direction held in 5 of 8 subjects (mean log ratio 0.17, p = 0.26), and median fitted values were 12.2 mm for HG and 13.4 mm for theta. The fit was poorly constrained because bipolar HG correlations were near their floor at the first measurable non-overlapping distance. Thus, in these data, bipolar HG envelopes from non-overlapping pairs were almost independent at about 7 mm spacing, but their spatial scale could not be resolved below that distance. Limits include one task segment per subject, one dataset, 8 subjects, clinical contact spacing (no distance below about 3.5 mm), linear correlation measures, no test of task-information redundancy, an undocumented recording reference, and unstable space-constant fits when correlations were near zero.

stereoelectroencephalography · SEEG · high gamma · theta · bipolar reference · volume conduction · brain-computer interface · spatial correlation · participation ratio