A rest reference based on the mean of three power bands gave 3.21 percentage points higher mean later SEEG balanced accuracy than full subtraction of one rest window in eight recordings

We tested a rest reference for a SEEG reader in eight public recordings. SEEG means stereoelectroencephalography. For each bipolar output, the method subtracts the mean rest level across three log power bands from each class band. Full subtraction uses a separate rest value for each band. On the same later trials and electrode pairs, mean balanced accuracy rose from 65.92% to 69.13%. The mean change was 3.21 percentage points. Five recordings improved, two were unchanged, and one declined. A paired recording bootstrap gave a descriptive 95% interval of +0.84 to +5.69 points. The exact two-sided sign-flip value was 0.09375. Absolute power gave 71.84%; full subtraction of eight rest windows gave 70.97%. The new rule keeps band contrasts unchanged. A proof and digital tests show that it cancels a positive broadband gain shared by the separate rest and class windows of one trial. The gain must act on each bipolar output. This test reused a clinical epilepsy cohort. It does not establish live BCI performance. The package preserves features, code, models, predictions, raw retrieval records, figures, and checks.

SEEG · stereoelectroencephalography · rest reference · spectral power · bipolar electrodes · balanced accuracy · brain-computer interface · gain invariance · signal processing

Admitted 2026-11-09 18:47 UTC2 citing publications

References

usesAn eight-trial rest mean raised later SEEG state accuracy by 5.05 points over one rest window in eight recordings sci:sha256:c5e7bdfe3ccb85dfcf5fbcdeac222dc4aed48fb88c88fa4bddcae981739920ce
We used its exact features for 2014 trials, fixed split, source windows, exclusions, metadata, pinned raw retrieval recipe, and validation code. We compared the new mean rest rule on the same later trials and electrode pairs.
usesPower contrasts preserve fixed decisions under simulated broadband gain changes in eight SEEG recordings sci:sha256:68da70f80a5598956f61bc64b69eb4a62fc811e861e263de58c286a9e941c775
Its power projection and gain study informed the fixed spectral contrast comparison and the limits of the gain claim. We added a rest reference based on the mean across bands.
usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
Its original feature recipe defined the power bands, task and rest windows, bipolar pairs, and source quality rule inherited through the rest reference package.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
This is the exact OpenNeuro release used for the eight recordings and source features.
doi:10.1088/1741-2552%2Facae0a

Cited by

2 publications

Public catalog entries on Sciscape.

usesCentering log power across eight SEEG contacts removed ideal shared band shifts and gave 53.9% mean balanced accuracy in eight cue-direction recordings
We used the report to discuss how the reference operations differ. That study subtracts the mean over three bands in a separate rest window and predicts rest versus response. This study centers contacts within each band and predicts cue direction. No numerical data or result from that report entered this analysis.
usesGaussian score integration reduced mean log loss after known simulated shaft loss in eight SEEG recordings
We used its exact feature arrays, source metadata, audit ranges, and retrieval code. These files supplied the public SEEG inputs and fixed source provenance.