Repeated contacts gave higher mean AUROC than unseen patients for a fixed five-neighbor rule on 30-second SEEG spectra from 31 patients
A fixed five-neighbor classifier used 30-second SEEG spectra from two run files in 31 patients. The test included 2,879 contacts and 271 clinical seizure onset zone contacts. Mean patient AUROC was 0.647 with repeated contacts, 0.607 with unseen contacts, and 0.583 with unseen patients. The repeated-contact minus unseen-patient difference was 0.064. A descriptive patient resampling interval was 0.036 to 0.098. In 999 controls, labels were shuffled within each patient. Each contact retained the same changed label across runs. Mean control AUROC was 0.551 with repeated contacts and 0.499 with unseen patients. The difference remained in all planned checks. Removal of overall log power reduced it to 0.022. These results show that access to repeated contacts can raise scores without clinical label meaning. The package preserves input spectra, pinned signal retrieval recipes, source code, neighbor maps, predictions, controls, figures, and repeat instructions. Independent checks and a second complete analysis passed. The result applies to this fixed method and selected short samples from OpenNeuro ds004100 version 1.1.3. It does not establish clinical localization performance or long-term contact stability.
SEEG · intracranial EEG · evaluation design · patient holdout · contact identity · arbitrary label controls · AUROC · epilepsy
References
5- usesIn 30-second HUP SEEG samples, reference changes reduced mean 30–100 Hz slope rank agreement more than run changes sci:sha256:b9c7b4646dba303292f492835f91dbb8026d4f35fcce938519d1d3fa6bcbff63
This study used the saved SEEG power spectra, source metadata, signal quality rules, and pinned retrieval manifests. They enabled the three training rules and the arbitrary contact label controls. The full input archive is preserved in source_archive.
- doi:10.1093/gigascience%2Fgix019
- doi:10.1038/s41598-024-56990-7
- doi:10.1038/s41598-025-02679-4
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The study uses the selected SEEG subset from ds004100 version 1.1.3, including its source recordings and clinical contact labels.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesAdding 30–100 Hz spectrum curvature raised mean SOZ AUROC from 0.598 to 0.703 in patient holdout tests of 31 HUP SEEG patients
The repeated-contact classifier result motivated exclusion of every test patient from training. This study applies that split rule to fixed slope and curvature models.
- usesFive spectral bands increased mean AUROC for SOZ labels beyond contact numbers in patient holdout tests of 30-second SEEG from 31 patients
This investigation used the exact 31-patient cohort, clinical labels, source spectra, and archived metadata. The source archive is preserved without changes. They enabled the patient holdout comparison of spectral and contact-number models.