In 256 RAM SEEG patients, spectral similarity between bipolar pairs declined with distance over 35–50 mm

We measured spatial dependence in spectral features of bipolar stereo-electroencephalography (SEEG) pairs from 256 epilepsy patients (27,974 pairs). Using archived 0.5 Hz log10 spectra from one 60 s task segment per patient, we compared pairs with no shared contact across seven bins of Euclidean distance between MNI midpoints. Within-patient semivariance increased with distance for the 2–40 Hz aperiodic exponent, 4–13 Hz peak height, and 2–40 Hz log10 power (mean per-patient Spearman rho +0.69, +0.66, and +0.45; Holm-adjusted p=0.0003 for each). The distance at which mean semivariance reached 90% of its 60–120 mm value was 50 mm (95% bootstrap CI 35–70 mm) for exponent and peak height, and 35 mm (25–35 mm) for log10 power. At 0–10 mm, exponent semivariance was 0.45 times its 40–80 mm value. Within-patient pairs remained more similar than pairs from different patients at matched distances; for the exponent, the within/between semivariance ratio ranged from 0.33 to 0.65. These findings support using the patient as the unit of inference or modelling spatial correlation in pair-level tests; pairs 20 mm apart were not independent in this sample. Distances were measured in MNI space, not along the cortex. Anatomy and tissue type were not modelled, and some distance dependence may reflect anatomy rather than volume conduction. The analysis used one task segment per epilepsy patient and bipolar re-reference only. The estimated range depends on the planned 0.9 criterion and bin edges.

SEEG · intracranial EEG · spatial dependence · aperiodic exponent · semivariance · bipolar pairs · MNI distance · patient-level inference