Sample replacement before 40 Hz forward and backward filtering increased mean SEEG agreement across reversed contact orders by 0.0052 at 20 to 100 ms in eight patients

We tested whether the order of short sample replacement and filtering changes measured SEEG agreement. We used 460 pulses at 46 sites in eight patients from OpenNeuro ds005448 version 1.0.0. We replaced samples from -2 to 12 ms and applied a fourth-order 40 Hz low-pass filter forward and backward. Raw samples outside the replaced interval stayed fixed. At 20 to 100 ms, replacement before filtering increased equal-patient mean signed agreement from 0.14950 to 0.15472. The change was 0.00522. The patient bootstrap 95 percent interval was 0.00139 to 0.01013. Seven of eight patient effects were positive. The mean normalized waveform difference was 3.88 percent. The change was much smaller at 80 Hz and at 50 to 150 ms. The no-filter difference was exactly zero. The package preserves source code, pulse matrices, all tables, 16 raw examples, and a retrieval recipe with fixed source versions and hashes. The cohort was reused from an earlier study. Population inference assumes independent patients. Replacement can change neural voltage and artifact. Neither processing order supplies a clean signal standard.

SEEG · intracranial EEG · electrical stimulation · signal processing · filtering · sample replacement · waveform agreement · epilepsy