Same-shaft SEEG band-power drift is weak, and shaft re-centering does not reduce late offsets

In task-free SEEG recordings from one session in 24 awake patients, we tested whether band-power offsets that remained after removing a global shift were more correlated between bipolar pairs on the same shaft. We used archived features from OpenNeuro ds005624: 4-s windows sampled once per minute, with log10 Welch power in four bands. Same-shaft pairs without a shared contact had higher correlations than pairs on different shafts in all four bands and all 24 patients (median r = 0.211–0.251 at 4–30 Hz and 0.095 at 70–150 Hz; sign-flip p = 0.00005 for each band). A 5-min moving mean gave a similar result at 70–150 Hz (r = 0.157 vs -0.002; 24/24, p = 0.00005), while shuffled shaft labels showed no structure. The shared component was small. Label-free shaft re-centering did not reduce late offsets in any band; for 70–150 Hz it increased median late mean z² from 2.25 to 2.62, with larger offsets in 21 of 24 patients (p = 0.0001). The analysis was exploratory and task-free, used one session per patient, did not measure decoder accuracy, and could not separate shaft effects from distance effects. The source data were read from a live OpenNeuro S3 URL; the exact dataset release was not established.

SEEG · band power · signal drift · shaft correlation · calibration · re-centering · intracranial EEG · OpenNeuro