Offline filtering reduced 70–110 Hz power change error against native SEEG in two-second windows from 35 records converted to 256 Hz

We tested a fixed anti-alias filter on 35 public motor-stimulation SEEG records from 21 patients in OpenNeuro ds006519 version 1.0.0. At 256 Hz, mean power change error fell from 0.551869 dB with direct sample selection to 0.000181 dB with filtering. The feature used 70–110 Hz power in two-second windows. Each record used channel medians. Each patient had equal weight. All 21 patients showed lower error. The paired reduction was 0.551688 dB. Its descriptive 95 percent patient bootstrap interval was 0.463518 to 0.645850 dB. The other planned checks also showed lower error. The package preserves code, spectra, source hashes, exact windows, figures, and retrieval instructions. Independent checks and a full source rebuild passed. It includes a recorded window-length amendment after a failed source-duration check. This offline test uses stored native SEEG as the reference. It does not establish decoding accuracy, artifact removal, or safe online sensing.

SEEG · signal processing · anti-alias filtering · resampling · power spectral density · brain-computer interface · reproducibility

Admitted 2026-11-09 18:53 UTC2 citing publications

References

usesA fixed detector found periodic voltage runs in 31 of 35 SEEG records from ds006519 version 1.0.0 sci:sha256:ea4058aa1dce526e987cca14e771bbeab34ccf9025624ac6983dbb7c42876d8e
Uses the source manifest, retrieval code, source metadata, and event-field audit for ds006519 version 1.0.0. The new study calculates power features from the full source voltages. It does not use the prior timing detector or its outputs to select records, channels, or windows.
usesdoi:10.18112/openneuro.ds006519.v1.0.0
This study uses all 35 released records from OpenNeuro ds006519 version 1.0.0.
doi:10.1016/j.neuroimage.2026.121822

Cited by

2 publications

Public catalog entries on Sciscape.

usesA 60 dB FIR design kept 500 ms, 70–108 Hz SEEG power within 5% after 4096-to-256 Hz reduction in 18 people
Its offline 80 dB rate-conversion study and report of a zero-power channel informed the amendment before current human-data scores. We added a fixed 80 dB comparison and a nonpositive-power rule. The new result tests shorter 500 ms frames and a 60 dB design with explicit causal delay. No prior score table or code was used to compute current features.
usesMedian 70–110 Hz power reduced large simulated fault error and brief 90 Hz probe response in two-second SEEG windows from 21 patients
Used the source manifest, metadata, retrieval code, and two-second window rule. Used its archived native signal examples and spectra to check the new coefficient extraction. The present study calculated new fault and signal probe results on the same released cohort.