Packed SEEG signs reduced float32 waveform bytes by 96.9 percent and mean response AUC by 0.028 in eight sessions at 20 to 500 ms
This study tested one bit per centered SEEG sample for pulse-response reading. It used 86 sites and eight sessions from OpenNeuro ds004696 v1.0.1. The first five pulses supplied each template. Five later pulses tested it. At 20 to 500 ms, mean sign AUC was 0.911. Mean full-voltage AUC was 0.939. The paired difference was -0.0284. Its patient bootstrap 95 percent interval was -0.0517 to -0.0097. Packed signs used 123 bytes per waveform, compared with 3,932 bytes for float32. This reduced waveform payload by 96.9 percent. Mean shifted-sign AUC was 0.551. At 200 to 500 ms, mean sign AUC was 0.814. All readers kept the same voltage-selected pair. Positive constant gain leaves the fixed-pair sign score unchanged in exact arithmetic. This result applies after voltage centering. It does not prove live control or neural origin. The archive contains packed codes, native trial waveforms, source checks, results, and three figures. GPL-licensed scripts are omitted from the SCI payload, so replay is not available from this package.
stereoelectroencephalography · SEEG · binary waveform coding · evoked responses · area under the curve · brain-computer interface
References
8- usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
The study keeps the pair eligibility rule and the first-five and last-five pulse split.
- usesFive calibration pulses gave mean later signed template AUC of 0.939 compared with 0.859 for one pulse in eight SEEG sessions sci:sha256:1d2aa103275ad7400befec4c899740ede8f2eb59a09897bdc3f4d4e190e54418
The source retrieval script adapts the range downloader and full-file hash check.
- usesFive-pulse signed templates did not show a clear AUC gain over energy on energy-selected pairs in eight epilepsy SEEG sessions sci:sha256:1406cee5468018906da44ece44823d82527f4a950f45c030e4c2416839f1b976
The full-voltage signed template supplies the main reader comparison.
- usesPair selection at 200 to 500 ms did not show a clear SEEG AUC gain with five calibration pulses in eight sessions sci:sha256:a29711f8fbecb6dcba14e964a02fa850d94a8ca01945702908f7cf4d85b6127b
The study uses native waveforms and calibration products from this archive. It keeps the full-window selected pairs.
- usesFour-bit storage reduced waveform payload by 87% and increased mean normalized 20–500 ms forecast error by 0.70% compared with float32 in eight SEEG sessions sci:sha256:1e8bc16598da030977cb3c6aa2ae9c6810a604515218a1685994c7fde5e81623
This earlier forecast study informed the scope distinction between model memory and binary waveform input.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
Source dataset version used for the selected native waveforms and source metadata.
- doi:10.1523/JNEUROSCI.2201-22.2023
- doi:10.1109/10.1386
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesTwo-bit rank codes raised later pulse response AUC by 0.019 over one-bit rank codes after five fit pulses in eight SEEG sessions
This study uses the fixed eight-person candidate cohort, chronological five-fit/five-test pulse split, native selected bipolar waveforms, fit-only pair selection energies, baseline reader code, and hash-checked source retrieval recipe. It compares new one-to-four-bit sample rank readers with the preserved sign and voltage readers. It does not add an independent cohort.
- usesTwo-bit rank codes raised later pulse response AUC by 0.019 over one-bit rank codes after five fit pulses in eight SEEG sessions
This study uses the fixed eight-person candidate cohort, chronological five-fit/five-test pulse split, native selected bipolar waveforms, fit-only pair selection energies, baseline reader code, and hash-checked source retrieval recipe. It compares new one-to-four-bit sample rank readers with the preserved sign and voltage readers. It does not add an independent cohort.