A 500-sample rest reference gave lower mean later-block SEEG balanced accuracy with 32 calibration trials in six patients
A separate 500-sample rest reference gave lower mean balanced accuracy on later SEEG blocks in six patients. We trained a fixed ridge decoder on 32 selected trials from the first block of OpenNeuro ds004473, version 1.0.1. Mean accuracy was 67.4% with absolute log power and 64.9% after subtraction of an earlier rest log-power vector. Five patients had lower accuracy and one tied. The exact patient sign-flip test gave p = 0.0625, so this small sample does not establish a population difference. Smaller calibration sets and control tests are preserved. An exploratory check found that subtraction preserved the paired class mean difference but increased median within-class feature variance by about 85% to 93%. All 849 archived trials were rebuilt from pinned public raw-data fragments; all feature arrays matched exactly. The package preserves code, source records, trial features, models, test scores, figures, and reproduction instructions. Tests used selected same-session windows and known response times. They do not establish live BCI control or a stimulation effect.
stereo-EEG · brain-computer interfaces · power reference · balanced accuracy · ridge regression · calibration · OpenNeuro
References
3- usesOffline SEEG power before a response separated task and rest across cue durations in six patients sci:sha256:6782e3a90eec43bc84b1aafb52750bb77fe4c10d7a324ea5f4f3e05395355d4c
This publication supplied the exact SEEG power arrays, trial selection records, fixed source metadata, EDF headers, pinned S3 range hashes, and extraction code used to test a separate rest power reference on strictly later blocks. All 849 feature trials were rebuilt and matched exactly. The recordings are the same cohort.
- usesdoi:10.18112/openneuro.ds004473.v1.0.1
The study used OpenNeuro ds004473 version 1.0.1; the version-pinned metadata records this dataset DOI and CC0 license.
- doi:10.1088/1741-2552%2Facae0a
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesThirty-two spread calibration trials gave lower mean later balanced accuracy than all early trials in eight SEEG recordings
The prior rest-reference and smaller-budget tests supplied calibration context. The new test compares 32 spread trials with all early trials in eight recordings. It uses a fixed mean-loss penalty and new pair selection for each set. No feature array or source code from the prior reference test was used.
- usesAn eight-trial rest mean raised later SEEG state accuracy by 5.05 points over one rest window in eight recordings
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.