Rest calibration reduces mean task accuracy in fixed SEEG imagery tests from two people
We tested SEEG readers with eight inputs on separate imagined grasp blocks from two people. Mean balanced accuracy was 66.25% after training on actual movement. A feature shift from imagery rest samples reduced it to 60.51%. A reader trained with imagery task labels reached 61.25%. The rest shift left ROC area unchanged, as expected for a constant shift of a linear score. The test used 86 paired imagery trials across four person and grasp cells. Visual cues and unknown contact identities limit the claim to task-period decoding. The package has code, features for all inputs, raw audit windows, source receipts, models, scores and four figures.
stereoelectroencephalography · motor imagery · brain-computer interface · task decoding · rest calibration · logistic regression · signal processing
References
4- 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 adapted its training-only signal input ranking and small logistic reader design. This enabled a fixed reader with eight selected amplifier inputs for the new actual-to-imagined grasp test. We used new source data, four power bands and separate test blocks.
- usesdoi:10.82901/nemar.nm000397
Source data: NEMAR nm000397, version 1.0.0, the BrainVision conversion used for the 30 analyzed blocks.
- usesdoi:10.5061/dryad.nq4fs
Original Dryad source deposit for the NEMAR conversion used in this study.
- supportsdoi:10.1371/journal.pone.0150359
Source study describing the executed and imagined grasp tasks and the clinical SEEG recordings used by the source dataset.
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesFixed early SEEG readers with eight inputs had 40% mean balanced accuracy for three force target ranks across executed grasps in four people
Uses the A/B feature extraction and source records inherited through publication 2526d00e-a871-4f0c-b786-e96c8ccd8647. The package preserves the original license notice. This work enabled the four-person feature set used for the new target test.
- usesEarly linear SEEG corrections increased mean relative squared error by 66.5% over cue and early sensor forecasts during grasps in four people
The inherited extraction method and A/B features descend from this work. This package preserves its source and license notice through the direct feature parent.
- usesMixed training reduced the larger early or late test Brier error from 0.325 to 0.255 with equal window counts in SEEG data from four people
This ancestor supplied the source extraction method and initial feature and raw audit format. We inherited that method through the later packages and used it to reconstruct the public records. The inherited code and data license notices are preserved.
- usesFixed SEEG readers scored 80% mean balanced accuracy early and 54% later across executed grasps in four people
Reuses 16 execution feature files, four raw audit files, metadata and source receipts for A and B. Adapts its finite-window extraction method for C and D. The new study tests transfer between executed grasps in all four people.
- usesTraining on late grasp windows raised mean late test balanced accuracy from 53% to 64% with eight SEEG inputs in four people
The direct parent inherited the extraction method and A/B features from this study. This package retains that code and data lineage and its license terms.
- usesPaired rest power did not show a clear three-target SEEG gain from actual to imagined grasp with eight inputs in two people
The earlier publication supplied all 30 feature blocks, 434 retained trials, eight raw audit trials, source metadata, receipt hashes and the extraction recipe. These files enabled the new three-target test from actual to imagined grasp. We also repeated its binary task-versus-rest reader as a positive control.