Past rest subtraction gave 58.5% mean balanced accuracy compared with 61.2% for spectral contrasts in an offline state test on eight SEEG recordings with 32 fit trials
We tested two ways to remove changes in signal gain from public SEEG data. We used OpenNeuro ds004473 version 1.0.2 and preserved features from an earlier publication. Each reader used eight bipolar pairs and the first 32 trials for fitting. The test used 674 later trials from eight people. Past rest subtraction gave 58.5% mean balanced accuracy for pre-response and rest windows. Spectral contrasts gave 61.2%. The paired mean difference was -2.75 percentage points. A descriptive 95% person bootstrap interval was -5.11 to -0.27 points. The exact two-sided sign-flip reference was p = 0.086. Both methods canceled gains that were equal across bands and stayed fixed across the used windows. Only spectral contrasts also canceled gains that changed after the baseline. These tests used synthetic gains. The cohort was used in earlier studies. The result is exploratory and does not establish a general difference between methods. The package contains source data records, raw examples, all features, code, predictions, models, and figures.
stereoelectroencephalography · SEEG · spectral contrasts · baseline subtraction · signal gain · balanced accuracy · brain-computer interface
References
5- supersedesPast rest subtraction gave 58.5% mean balanced accuracy compared with 61.2% for spectral contrasts in an offline state test on eight SEEG recordings with 32 fit trials sci:sha256:351fefcac4162f0abb64adbe2c93265a970cc3899c64652f1c2287cfe11e12f3
This package corrects the file hash checker. It excludes the separate metadata folder that sci fetch adds during retrieval and documents this rule. The numerical analysis, saved predictions, and scientific results are the same.
- usesSpectral contrasts gave 61.2% mean balanced accuracy under fixed shaft gains within ±3 dB in eight SEEG recordings sci:sha256:14a3a6432c644bebc872ba12928232e79781ea1a740dff5445b1a5aabf94d042
We used the spectral contrast definition and the 32 trial comparison. We repeated its absolute full and spectral contrast mean scores. Its fixed shaft gain test led to the new comparison with a past baseline and gains that vary between trials or windows.
- 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 used the complete feature cache, public data retrieval code, trial records, metadata, and raw range records. These files made the new past rest comparison possible. We preserved their hashes and checked raw examples against the public source.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset: OpenNeuro ds004473 version 1.0.2, whose saved metadata identifies the analyzed data release and CC0 license.
- doi:10.1088/1741-2552%2Facae0a
Related study describing the public forced two-choice SEEG task data; this analysis uses different readers, windows, exclusions, and test split.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesCurrent rest subtraction gave no clear log loss gain in eight offline SEEG key readers after task changes with 64 calibration trials
The earlier state-label rest subtraction test motivates the present feature comparison. The present target is the response key, and its input eligibility uses the corrected cohort. The two tests have different scopes.