Causal filtering gave more output changes above baseline RMS at 50–80 ms with matched 70–110 Hz gain (five SEEG patients)

We tested how uncertain pulse samples affect later SEEG output when two filters have equal steady-state magnitude response. We used 400 pulses from five patients in OpenNeuro ds004457 version 1.0.2. We replaced samples from −2 to 10 ms with a straight line. The filters passed 70–110 Hz. One used forward-backward filtering. The other used two causal forward passes. At 50–80 ms, the output change exceeded baseline RMS in 174/400 causal outputs and 8/400 forward-backward outputs. The equal-patient increase was 41.50 percentage points. Its descriptive 95% patient bootstrap interval was 30.50–52.50 points. All five patient differences were positive. An added exact two-sided patient sign-flip reference gave p = 0.0625. A causal gamma guard set from four early pulses failed on 26/200 later pulses at the same sites. The two filters gave the same total effect energy within numerical precision on extended records. The main result concerns effect timing. Fresh raw retrieval and extraction matched every stored epoch. A repeat matched all main result files. The package preserves code, exact digital inputs, source hashes, all scores, delay tests, and four figures. This study extends an earlier audit in the same cohort. It does not identify clean neural voltage or a safe stimulation delay.

causal filtering · zero-phase filtering · stereoelectroencephalography · electrical stimulation · signal processing · pulse artifact