SEEG bipolar-pair spectra were reproducible across two FR1 sessions in patients with epilepsy

We tested if the 1-45 Hz spectrum of an SEEG bipolar pair is stable between two recording sessions. We used 60 s of FR1 task data from session 0 and session 1 for each patient. The plan was hashed before we read session-1 data. We matched 10567 pairs in 107 patients by channel name. Spectral shape correlated more with the same pair than with other pairs (mean D 0.035, 95 % CI 0.031-0.039, 103 of 107 patients, sign-flip p 1e-5). Top-1 identification accuracy was 0.28; chance was 0.019. The same pair also correlated more than the nearest non-overlapping pair at a median of 8.4 mm (D_nb 0.015, 101 of 104 patients). Approximately 42 % of the same-pair advantage was not shared with this neighbour. The alpha map rank correlation between sessions was approximately 0.78. Within patients, across-pair ICC was 0.68-0.94 for relative band power, aperiodic exponent and broadband power. Subject-level ICC was only 0.70 for broadband power and for the exponent. Limits: patients had epilepsy, data are task data, and the time between sessions was not checked. The package contains code, session-1 spectra, byte-level provenance, results and reproduction steps.

SEEG · intracranial EEG · spectral stability · bipolar pairs · cross-session reliability · FR1 · epilepsy