Trial baseline subtraction increased log power variance for most tested features in eight public SEEG recordings
We tested whether trial baseline subtraction improved response side classification in 2,030 trials from eight patients. The source was public SEEG data from OpenNeuro ds004473, version 1.0.1. Features were beta and gamma log10 power from adjacent contact pairs. The comparison used a one second response window and a separate one second baseline before fixation. A fixed ridge decoder used five consecutive test blocks and training data for standardization. The median corrected to absolute training variance ratio was 1.86 to 2.14 across patients. Mean balanced accuracy was 53.12% with absolute power and 50.07% with subtraction. The paired change was -3.05 percentage points, with a 95% patient bootstrap interval from -6.73 to 0.66 points. The exact sign flip test gave p = 0.1875. Thus, the test did not establish an accuracy decrease or increase. The package preserves source features, code, predictions, folds, inference draws and retrieval instructions. A fresh subset of 48 source epochs matched the preserved hashes and features exactly.
SEEG · stereoelectroencephalography · trial baseline subtraction · log power · spectral analysis · balanced accuracy · ridge classification · time block validation · OpenNeuro · beta band · gamma band
References
4- usesA fixed power decoder detects response side under time holdout in one of eight public SEEG recordings sci:sha256:a0fd1f7fd080c7eb6ba602955ec758d6637686847b98c863c9525c95eaceeaf5
We used the preserved beta and gamma log power features, labels, source locks and extraction code from this publication. These files enabled the paired trial baseline comparison. We also reproduced its absolute power test predictions.
- usesdoi:10.18112/openneuro.ds004473.v1.0.1
The analysis used the preserved features and metadata for OpenNeuro ds004473 version 1.0.1 and checked a subset against its fixed source objects.
- doi:10.1088/1741-2552%2Facae0a
- doi:10.3389/fpsyg.2011.00236
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesAdding current baseline power reduced mean prediction error for later response power by 0.075 percent in eight SEEG patients with the specified ridge rule
Used its 2,030 aligned trial feature maps, trial tables, source metadata, and unchanged EDF reader. These files supplied the fixed inputs and raw rebuild recipe for the new forward response power prediction test.
- usesAdding high gamma lowered loss for pre-key window pairs in fixed tests of later trials from eight SEEG recordings
We used its trial tables, original features, public metadata, pinned source objects, range hashes, and EDF reader. These files enabled fresh retrieval and validation of the new half-second windows.
- usesCoefficients fitted to earlier baselines raised later beta log power variance within task groups in eight ds004473 SEEG patients
We used its preserved feature inputs and source records. Its finding about fixed subtraction informed the new question. We tested a new outcome: transfer of fitted coefficients from earlier trials to later variance within task groups. We did not obtain new patients.