Choice among 12 scores gave a mean AUC gain of 0.011 in separate seizures from 29 ds004100 SEEG patients

We tested whether score choice with one seizure improved clinical onset mark ranking in a separate seizure. The input was public ds004100 version 1.1.3. We retained 29 patients and 471 matched contact pairs on 62 named shafts. Each patient chose among 12 power scores with run 01 marks. Mean run 02 AUC gain over fixed absolute bipolar gamma was 0.0108. The 95 percent patient bootstrap interval was -0.0309 to 0.0518. The exact two-sided sign-flip probability was 0.6274. Mean training gain was 0.0674. The main test did not establish a gain or equal performance. Population intervals require independent patients with exchangeable sampling. The sign-flip test also requires symmetric null effect signs. The study reused a cohort that had informed the method. It does not test new patients or clinical outcomes. The package preserves features, scores, choices, ties, patient effects, bootstrap draws, code, source versions, and repeat instructions.

stereoelectroencephalography · SEEG · epilepsy · seizure onset · spectral power · AUC · score selection · patient bootstrap

Admitted 2026-11-09 19:35 UTC13 citing publications

References

usesEarly bipolar gamma change had higher local order agreement than pre-onset change across two seizures in 31 ds004100 SEEG patients sci:sha256:4a1d907056065054a6e97e702e230129688ba671a410db016919665e38c09e05
Its study of local order repeatability in two seizures informed the separate-seizure question and patient sampling unit. Its numerical results are not input to the new score selection benchmark.
usesBipolar gamma power had mean clinical onset mark AUCs of 0.900 at 0 to 10 seconds and 0.883 at 10 to 20 seconds within named shafts in 26 ds004100 SEEG patients sci:sha256:907aa78467fe5bfb0a76cd59da5caa7cbfb46102973ed12c999a62fce1d8acdf
This study uses its 64 unchanged power files, pinned public metadata, versioned raw range recipes, and source code. Its same-cohort findings informed the fixed absolute bipolar gamma comparator.
usesEarly gamma ranks had higher mean agreement across references than across seizures in 32 ds004100 SEEG patients sci:sha256:4537e0031e79d6302fa623490378be55c0329af497da96e30a06a63f697a900a
Its inherited feature extractor and source range records are preserved. The new public-source check uses its transport helper and separately checks EDF decoding and spectral power.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
Source data: OpenNeuro ds004100 version 1.1.3. The analysis used the preserved source records and 64 unchanged power files.
doi:10.3389/fneur.2026.1862336

Cited by

13 publications

Public catalog entries on Sciscape.

usesCentered averaging let later voltage change computed gamma power before onset with causal filtering in 64 SEEG records
Used its 64-record selection, channel and event metadata, source object versions, and EDF layout. These inputs enabled new ten-second native voltage retrievals. New pairs were selected from metadata. Its derived power arrays and AUC results were not used.
usesA fixed gamma model gave mean within-shaft AUC 0.818 for SOZ marks in 14 SEEG patients excluded from fitting
We inherited its fixed 32-patient selection and EDF transport layout through the direct parent. We did not use its earlier power measurements.
usesOnset gamma rise and post-seizure gamma fall had mean cross-seizure rank correlation -0.033 within shaft and label groups in 29 SEEG patients
The fixed 32-patient SEEG cohort and runs 01 and 02 were inherited through the direct feature source. The preserved source selection audit comes from this cohort selection.
usesMean within-shaft SOZ AUC for early gamma change stayed above 0.5 across all opposite-mark pair labels in 29 SEEG patients with equal-mark labels fixed
This publication first generated the all-pair power arrays. Its extraction code is preserved and tested here.
usesA ten-second local baseline changed mean gamma threshold balanced accuracy by −0.16 percentage points in offline tests of 32 SEEG patients
We used the original power extraction code and its fixed EDF byte ranges to rebuild all 64 compact inputs from public SEEG recordings. We checked every reconstructed array and preserved the source version and range hashes.
usesA separate implementation reproduced the ten second calibration accuracy change across two seizures in 32 SEEG patients
Files and provenance from this earlier analysis are preserved in the direct source archive. They support the source chain and document reuse of the same public power data.
usesGamma thresholds set from five baseline windows flagged 101 of 320 baseline test windows after transfer and 81 with local use in 32 SEEG patients
This work supplied the original power extractor and separate NumPy EDF reader. Their verified source files enabled full retrieval and independent source checks.
usesAveraging gamma power before the logarithm reversed a mean 7.0 percent of SOZ comparisons within shafts 10–30 seconds after seizure end in 29 SEEG patients
We inherited its fixed 32-patient record selection, source metadata, EDF layout, and S3 version pins through the direct gamma study. Its power score selection results were not numerical inputs.
usesGamma decrease 10–30 seconds after seizure end gave mean within-shaft AUC 0.790 for SOZ marks in 29 ds004100 SEEG patients
This study supplied the fixed record list, metadata, EDF headers, and source version pins. EDF parsing and range checks adapt its code.
usesGamma thresholds set from five baseline windows flagged 101 of 320 baseline test windows after transfer and 81 with local use in 32 SEEG patients
This work supplied the original power extractor and separate NumPy EDF reader. Their verified source files enabled full retrieval and independent source checks.
usesA gamma pair selected in one seizure gave mean AUC 0.779 in a separate seizure compared with 0.807 for pooled SEEG in 32 patients
This archive supplied all 64 unchanged power files, fixed source metadata, S3 object versions, range hashes, extraction code, and raw examples. We used these files to test time classes across two seizures in 32 patients. Clinical contact classes and the source study results did not enter our selection or prediction rules.
usesThe 95% patient interval included zero AUC gain from added post-seizure gamma rank weights of 0% to 50% in separate seizures from 29 SEEG patients under independent, exchangeable patient sampling
We read this report before the local plan. Its score selection across separate seizures informed our fit and test design. The post-seizure parent supplied the actual input files.
usesAfter shaft and baseline control, the 95% interval for mean gamma change correlation included zero in 32 ds004100 SEEG patients. Patient sampling assumed independence and exchangeability.
We inherited the 32-patient source selection, clinical event rules, public source object pins, and EDF transport and decoding design through the gamma parent. The copied EDF helper identifies this code lineage.