An eight-trial rest mean raised later SEEG state accuracy by 5.05 points over one rest window in eight recordings

We tested a separate rest reference for reading task states from public SEEG data. The source was OpenNeuro ds004473 version 1.0.2. Early trials set eight bipolar pairs and separate logistic readers. The later test used 618 trials from eight recordings. A mean of current and seven earlier rest log powers gave 70.97% mean balanced accuracy. One rest window gave 65.92%. The paired change was +5.05 percentage points, with a patient bootstrap 95% interval of +2.18 to +8.20. Seven patients improved. Absolute power gave 71.84%. A proof and digital tests showed cancellation of fixed positive bipolar-output gains. The gain must be shared by both class windows and every reference window. Changing gain left a residual error. The eight-window reference spanned about 40 to 144 seconds. All feature arrays matched a fresh rebuild from pinned public raw ranges. Independent power and optimizer checks passed. A separate analysis gave exact numerical matches. The package preserves code, inputs, provenance, fitted readers, predictions, figures, and reproduction instructions. This is an exploratory offline test in a reused clinical cohort. It does not establish continuous BCI control or stimulation effects.

stereoelectroencephalography · SEEG · rest reference · balanced accuracy · log power · gain stability · brain-computer interface

Admitted 2026-11-09 18:21 UTC1 citing publication

References

usesA 500-sample rest reference gave lower mean later-block SEEG balanced accuracy with 32 calibration trials in six patients sci:sha256:9759feb614be9dfcf0bf5fb95eb58203efb8807933b43c94e772b1b341b406e8
The separate short-rest method and its measured variance cost set the research question. We tested whether a mean over eight separate rest log powers reduced that cost on the full eight-recording feature set.
usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
We used all eight feature records, source metadata, trial eligibility, and raw retrieval code. We rebuilt every feature array from pinned public raw ranges. We added a rolling rest mean and its later-trial comparison.
usesA correction from eight known rest windows kept 69.7% mean accuracy under simulated constant bipolar gains of 0.5 to 2 in eight SEEG recordings sci:sha256:547af54a5b0969526352c29c2e536bb43066e3531d395d0e0b67318818002e77
Its correction from eight known rest windows set the gain-stability question. We tested a reference that updates every eligible trial, proved fixed-gain cancellation, and measured residuals from defined gain changes.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis used all eight recordings from OpenNeuro ds004473 version 1.0.2; the saved dataset metadata records this exact version and a CC0 license.
doi:10.1088/1741-2552%2Facae0a

Cited by

1 publication

Public catalog entries on Sciscape.

usesA 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 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.