Common average 30–80 Hz spectral slopes had mean onset-label AUC of 0.473 in paired 60 s SEEG records from 36 patients

We tested whether a simple spectral slope gives stable contact ranks and identifies clinical seizure onset contacts. The source was OpenNeuro ds004100 version 1.1.3. We used paired 60 s interictal records from 36 patients and 4,260 matched contacts. The fixed positive onset score was a higher 30–80 Hz slope under common average reference. The fit removed 55–65 Hz and did not remove spectral peaks. The equal-patient mean onset-label AUC was 0.473. Its 95% patient bootstrap interval was 0.404–0.543. Median rank correlation between files was 0.879, with interval 0.839–0.912. All planned fit checks had mean AUC intervals that included 0.5. Thus, high median rank agreement did not establish above-chance onset discrimination for this score. The result does not prove equivalence to chance. The biological time gap between files is unknown. We downloaded all 72 raw excerpts and rebuilt their spectra. Every contact rank and AUC reproduced exactly. The package preserves spectra, source metadata, raw retrieval recipes, code, contact scores, patient results, figures, and checks. The internal plan was saved before slope calculation and has no external registration.

stereoelectroencephalography · spectral slope · seizure onset · rank agreement · area under the curve · common average reference · OpenNeuro

Admitted 2026-11-09 16:55 UTC3 citing publications

References

usesCommon average reference raises median relative gamma rank agreement in two 60 s SEEG records from each of 36 HUP patients sci:sha256:3f5f9f6ae779730a7bf70359ae9ddc56a5032c2e8abdb97c0cf3c127d6082e74
Uses its fixed 36 patient spectrum files, contact selection, source metadata, raw byte hashes, and retrieval and EDF reader code. These files enable the new fixed-direction slope, rank-agreement, and onset-label AUC analysis.
usesA 30–80 Hz SEEG slope score had median onset AUC 0.598 within named shafts under the supplied reference in 34 patients sci:sha256:b1540e7ea5e7e85e110d7fc95f44a4ca2b09fd269cadc931de598e8caf1f3f94
Uses this related result for the scope comparison in the report. It tests a steeper-slope score within named shafts under the supplied reference. Our test uses a flatter-slope score across all retained contacts under common average reference. Both analyses share the parent spectra and do not provide independent cohort confirmation.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analysis uses the version 1.1.3 OpenNeuro ds004100 snapshot for source data and clinical onset labels.
doi:10.1038/s41593-020-00744-x

Cited by

3 publications

Public catalog entries on Sciscape.

usesBipolar reference gave higher mean onset AUC than CAR for 30 to 80 Hz SEEG pair slopes within shafts in 31 patients with stable, matching contact marks
Supplied the earlier CAR contact rank result. Its different score direction, contact unit, and across-shaft pool helped define the new paired reference question and its scope.
usesMean bipolar reference effects exceeded file differences for 30–80 Hz slopes in paired 60 s SEEG records from 36 patients
The prior CAR-only slope analysis of this HUP cohort informed the study scope. This study adds paired bipolar spectra and a reference-versus-file contrast. We did not reuse its power values or test clinical labels.
usesPower averaging order gave positive and negative 30–80 Hz exponent changes under common average reference in paired SEEG records from 36 patients
This study uses its fixed 36 patient cohort, matched contact list, source metadata, frozen raw byte ranges, and EDF reader. Fresh raw signals produce the new epoch spectra and the test of power averaging order.