A fixed SEEG decoder identified key side before keypresses in one of eight patients under conditional label tests
We tested key side in 2,030 trials from eight patients in public SEEG dataset ds004473, version 1.0.1. Each window contained 500 samples, or 0.500294 seconds, before or after the registered keypress. A fixed power decoder used five consecutive time block tests. Only participant 2 passed the before-window label test after correction across 16 patient and window tests. Its balanced accuracy was 76.69 percent, with adjusted p = 0.008. The mean accuracy was 53.34 percent before and 53.15 percent after the keypress. The mean after-minus-before difference was -0.18 percentage points. Its patient bootstrap 95 percent interval was -5.04 to 5.24 points; the exact sign-flip p was 0.96094. This does not establish equal accuracy. The label tests assume exchangeable labels within each time block and fast or slow instruction. Cue side and movement before the keypress can contribute to the signal. The analysis does not show prediction before movement onset or online use. The package preserves code, power features, trial labels, predictions, null results, and pinned source retrieval. Independent checks and complete offline and raw rebuilds passed.
stereoelectroencephalography · SEEG · movement decoding · key side · power features · label permutation · time-window analysis
References
3- usesA fixed power decoder detects response side under time holdout in one of eight public SEEG recordings sci:sha256:a0fd1f7fd080c7eb6ba602955ec758d6637686847b98c863c9525c95eaceeaf5
We reused its source metadata, event and behavior join, EDF byte reader, and fixed ridge routines. We knew its full response-window results. The present study newly extracted equal windows before and after the keypress in the same eight recordings. It is not an independent data replication.
- usesdoi:10.18112/openneuro.ds004473.v1.0.1
Source dataset ds004473, version 1.0.1, supplied the SEEG recordings and metadata used in this analysis; its preserved dataset metadata states CC0.
- doi:10.1088/1741-2552%2Facae0a
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesTests with separate blocks did not detect greater late beta power above baseline at bipolar SEEG pairs selected for decreases before keypresses in eight patients
This publication supplied aligned trial records, before and early-after power arrays, metadata, pinned raw object versions, source code, and raw examples. These files enabled the new baseline and late power measurements and the separate-block test.
- usesSEEG power before keypresses identified response keys across timing instructions in one of eight patients under conditional tests
We reused its before-keypress power features, trial tables, pinned signal ranges, fixtures, and extraction code. We knew its patient 2 result. We checked the source signals again and tested instruction transfer with matched training counts.
- usesAdding beta phase changed SEEG key side prediction by -0.19 percentage points in a fixed model tested on later trials from eight patients
We read the earlier power decoder result to choose a separate paired power and phase question. Its windows, sample, and split differ. We did not reuse its predictions or treat its scores as a direct comparison.
- usesMean SEEG key decoding across task conditions exceeded chance in eight patients under conditional label exchange
We used the feature arrays, trial labels, source metadata, raw fixtures, and extraction code. They supplied the fixed cohort for the new task-condition transfer test. Fresh public ranges reproduced the features exactly.