Mixed windows gave 9.1 percentage points higher mean balanced accuracy than separate records for a four-pair seizure reader in 28 SEEG subjects
We tested a small SEEG reader with two seizure records per subject from OpenNeuro ds004100 version 1.1.3. Every split used 40 training windows and 40 test windows. Mean balanced accuracy was 88.1% with mixed windows and 79.0% with a separate test record. The mean difference was 9.1 percentage points. Its 95% subject bootstrap interval was 5.2 to 13.4 points. Contiguous window halves gave 80.8%. These results apply to this fixed reader and 28 selected subjects. They do not give a long-term false alert rate or a measured benefit from stimulation. The package preserves features, source metadata and headers, raw audit windows, source range recipes, all fitted models, code, checks, and figures.
stereoelectroencephalography · seizure detection · brain-computer interface · validation · record transfer · balanced accuracy
References
6- usesRandom splits gave higher mean speech and music block accuracy with eight contact pairs in six SEEG recordings of one film sci:sha256:072f5aa938e2b517feb5005f8060dfbfa31bf4aa3e4a118c8dd327af55fd776f
Its matched-count comparison of mixed windows and separate film sections informed the validation design. This study tests separate seizure records with a different fixed reader.
- usesLocal calibration lowered mean pre-onset window alerts from 25.6% to 8.8% for a median monitor in 28 SEEG record pairs after onset shafts were excluded sci:sha256:3f75c11b0a1b56a89e441827aad7b6b86812d0548cc966f8a1c9eea17cc7bd0e
Its result on calibration changes between seizure records informed the record-transfer question. This study retrieved source data again and used new code and features.
- usesRandom window splits gave 50.9% mean balanced accuracy for random trial labels in eight SEEG recordings with no shared samples sci:sha256:b17d0614217cf4800acad6818885eba77de595c8adca2dafa8145070f3865ba2
Its random-label test with disjoint SEEG windows informed the rule that retained windows must share no source samples. This study measures a seizure-state validation gap.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
OpenNeuro ds004100 version 1.1.3 supplied the HUP iEEG source data and metadata analyzed here.
- usesdoi:10.21105/joss.01896
The source dataset README states that MNE-BIDS was used to convert the dataset into BIDS format and provides this citation.
- doi:10.1038/s41597-019-0105-7
The source dataset README cites the iEEG-BIDS paper for the intracranial electrophysiology data format.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesA 10-second score offset gave lower mean balanced accuracy for a four-pair seizure reader in 28 SEEG subjects
We reused its exact 28-subject cohort, shared pair lists, test features, source metadata, EDF utilities, and reader rules. We added early calibration data and reproduced its 56 separate-record fits.
- usesTen baseline seconds reduced mean baseline alerts and seizure sensitivity for a fixed four-pair reader in 28 SEEG subjects
This study reuses the 28-subject cohort, paired-record features, exact four-pair reader, EDF audit samples, source metadata, and pinned provenance. It adds fixed baseline-only calibration tests and fresh checks of all 56 source ranges.
- usesEarly baseline correction lowered mean balanced accuracy for a fixed SEEG reader in 28 people
We reused its 28-person cohort, 56 scored feature archives, fixed four-pair reader, and pinned source recipes. We added new early calibration data and a fixed threshold correction test. All uncorrected fitted arrays reproduced exactly.