Lower contact numbers ranked source onset marks with mean AUC 0.671 in 36 SEEG patients
We tested contact number as a baseline for seizure onset ranking in public SEEG data. The source was OpenNeuro ds004100, version 1.1.3. The test used 4,252 contacts from 36 patients. A missing source onset mark defined the comparison class. This assumes complete source marks. Lower numbers had mean AUC 0.671, with a 95 percent patient interval of 0.605 to 0.732. A fixed low frequency EEG score had mean AUC 0.647. Its paired difference was -0.024, with an interval of -0.107 to 0.067. The primary result did not establish a difference or equivalence. Within mixed name groups, the number baseline had higher mean AUC. A model with both inputs had held-out mean AUC 0.715. Its mean gain over number was 0.044. That secondary interval held predictions fixed and omitted model training uncertainty. Contact names did not establish physical depth or implant shape. The package preserves complete source inputs, code, patient results, model folds, pinned raw crops, and reproduction instructions. Independent checks passed. A complete rerun reproduced 16 numerical files byte for byte. The study does not test treatment benefit.
stereo-electroencephalography · seizure onset marks · contact number · area under the curve · patient-held-out models · low-frequency EEG · spatial bias
References
5- usesAverage references reduced low frequency seizure onset ranking AUC in two short SEEG recordings from each of 36 patients sci:sha256:ac5e50a0fe3911b4b562b31f3ee7fb2ab9caf8fe9695b3ad882f2c66ce6e9f33
Uses all 72 archived window-power files, exact contact mask, source channel metadata, extraction code, and pinned retrieval records. The complete input publication is copied and verified. All 108 patient-by-reference signal rows are reproduced. The new tests compare contact number with the fixed signal and use patient-held-out rank models.
- usesA low frequency score had 0.058 excess mean AUC under uniform shaft label permutations in 72 short SEEG records from 36 patients sci:sha256:739286ddbdfce424bcb3b92db5bc54ead8ee8a7bb7c626601848a9a283369a2c
Uses the related study for scientific context after the new calculations. Its known-label group null and this unknown-label contact number predictor answer different questions. No numerical results, inputs, or model settings were taken from it.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The archived feature set and source metadata used in this analysis come from the HUP iEEG Epilepsy Dataset, OpenNeuro ds004100 version 1.1.3.
- doi:10.1088/1741-2552%2Fac90ed
- doi:10.1038/s41598-019-42347-y
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesA fixed logistic model had lower mean seizure onset ranking AUC with two added power variation features in 18 SEEG test patients
Its contact number result motivated the two contact name inputs. These inputs form a simple baseline before median power and power variation are added.
- usesA fixed low frequency score added 0.034 to mean onset-label AUC above a coordinate and number model in 36 SEEG patient holdouts
The prior contact-number result motivated a fixed contact-number control. This study calculates its own contact-number input from source names and adds a coordinate model and patient bootstrap model refits. It does not reuse prior numerical predictions.