Trial baseline subtraction gave an uncertain AUC gain in 14 offline eight-channel SEEG tests at penalty 10
We tested whether an additional baseline for each trial improves choice reading from public MetaRDK SEEG signals. We used eight selected bipolar channels per model, early fitting trials, and 1,167 later test trials from 14 people. At penalty 10, mean choice AUC within visual directions was 0.4936 for level input and 0.5059 for level minus baseline. The paired change was +0.0123. Its descriptive 95% person bootstrap interval was -0.0322 to +0.0619. Seven people improved. The exact sign-flip p value was 0.6345. The evidence does not establish a gain or equal performance. Subtraction changed 5.5 selected channels on average. Fixed-channel checks also did not establish a gain. The package preserves compact data, 210 saved models, predictions, code, three figures, and pinned source recipes. Two full public-source rebuilds matched every saved array. This is an exploratory offline test in a reused cohort. Supplied smoothing and retrospective event times prevent a live timing claim. We did not test stimulation.
stereoelectroencephalography · brain-computer interface · choice decoding · baseline subtraction · high-gamma activity · AUC · channel selection · logistic regression · random dot motion · reproducibility
References
3- usesOffline eight-channel SEEG models gave 0.484 mean choice AUC within visual directions in 8 new MetaRDK participants sci:sha256:000ec8109a497514631b4d12ae086e5a3ac4edce34c56551ee0604fb598fe426
We reused all 14 compact feature files, trial audits, extraction code, source range reader, and fitting rules. We added a trial baseline comparison and fixed-channel checks. We rebuilt source arrays for people 01 and 02 in this run.
- usesdoi:10.18112/openneuro.ds006253.v1.0.3
The analyzed processed SEEG signals and behavior records come from this exact MetaRDK release.
- supportsdoi:10.1038/s41467-025-61744-8
This source study documents the task and upstream signal processing described in the report.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesSignal subtraction reduced mean AUC on later errors by 0.1205 in four offline SEEG readers fit on correct trials with eight channels at penalty 10
We used its 14 compact signal datasets, extraction rules, source provenance, and numerical fitting code. We changed the training rule and tested movement subtraction on later errors.
- usesSEEG readers fitted after movement lost 0.085 mean choice AUC before movement in 14 offline tests with eight channels at penalty 10
This publication supplied all 14 compact feature files, source audits, metadata, trial rules, and source rebuild code. These inputs enabled the frozen-reader transfer test.
- usesSignal subtraction reduced mean AUC on later errors by 0.1205 in four offline SEEG readers fit on correct trials with eight channels at penalty 10
We used its 14 compact signal datasets, extraction rules, source provenance, and numerical fitting code. We changed the training rule and tested movement subtraction on later errors.