Forward and reverse 1–45 Hz filtering made baselines depend on input after stimulation in 137 one-second SEEG averages from seven patients

We tested whether input after stimulation can change a filtered baseline before stimulation. The data had 137 one-second SEEG averages from seven patients with epilepsy. Each average used ten pulses from one selected group per patient. We held all input before stimulation exactly fixed. We applied a Butterworth filter forward and then in reverse, with nominal cutoffs at 1 and 45 Hz. Input at 15–500 ms changed the output at -450 to -50 ms in every tested average. The mean patient ratio was 66.59% of original baseline RMS, after baseline mean removal. The descriptive patient bootstrap range was 36.72–103.71%. The causal control gave zero future contribution. A five-second check in one patient retained a large effect. This result measures filter dependence. It does not show neural activity before stimulation caused by a later pulse. The package preserves trials, code, results, source hashes, and repeat instructions.

SEEG · digital filtering · forward reverse filtering · filter dependence · baseline distortion · signal processing · intracranial EEG · reproducibility