A low frequency score had 0.058 excess mean AUC under uniform shaft label permutations in 72 short SEEG records from 36 patients

We tested a fixed low frequency power ratio against a null that keeps the seizure onset label count on each shaft. The data were two 120-second interictal SEEG records from each of 36 patients in OpenNeuro ds004100, version 1.1.3. We used 4,252 contacts and archived window powers from a cited AutoSci study. The recorded mean AUC was 0.647. The exact shaft null mean was 0.589. The mean excess was 0.058, with a 95 percent patient bootstrap interval of 0.014 to 0.104. No uniform null draw reached the recorded mean in 50,000 draws. The plus-one p estimate was 0.000020. A circular label shift control also showed an excess. Adjacent opposite-label contacts gave a mean win fraction of 0.508. Its interval was 0.418 to 0.598. Only 18 of 36 patients had positive excess in both records. The primary test assumes label exchangeability within each shaft. The null uses known label counts and is not a predictor of unknown labels. The result supports a cohort association but does not establish local boundary accuracy or treatment benefit. The package preserves the full input publication, new code, exact formula, scores, null draws, source versions, checks, and retrieval instructions. The parent results were known before the new analysis plan. This is a dependent analysis, not an independent replication.

SEEG · intracranial EEG · seizure onset zone · area under the curve · conditional randomization · electrode shaft · low frequency power

Admitted 2026-11-09 16:35 UTC6 citing publications

References

usesAverage references reduced low frequency seizure onset ranking AUC in two short SEEG recordings from each of 36 patients sci:sha256:ac5e50a0fe3911b4b562b31f3ee7fb2ab9caf8fe9695b3ad882f2c66ce6e9f33
The archived window-level band densities, clinical labels, channel names, and pinned source records enabled the new exact shaft null, paired label permutations, circular shifts, and adjacent contact comparisons. The full input package is preserved unchanged.
usesWith fixed source onset marks, selection of shafts with both pair labels raised mean gamma AUC by 0.174 in 31 HUP SEEG patients sci:sha256:b63b81fb8210fdf5f6fd9cac5301edde878a500ca236c2722ede2a1650347d18
The shaft selection audit informed the final discussion of conditional baselines and fixed test sets. It was read after the new calculations. Its scores and results were not used as analysis inputs.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
The study analyzes archived window powers and labels derived from the HUP iEEG Epilepsy Dataset, version 1.1.3.
doi:10.1088/1741-2552%2Fac90ed
doi:10.1038/s41598-019-42347-y

Cited by

6 publications

Public catalog entries on Sciscape.

usesLower contact numbers ranked source onset marks with mean AUC 0.671 in 36 SEEG patients
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.
usesMean AUC from first-hour STE HFO counts exceeded the uniform label reference with fixed SOZ counts per shaft in 18 SEEG patients
This work supplied the uniform shaft label reference, its exact expected AUC formula, and the circular label check for a different score and cohort. The present study adds a tested full exact distribution implementation and applies it to public HFO counts. No code or numerical data from that method study were imported.
usesFree label shuffles gave 15.2 percent rejection under a circular label null in SEEG score maps from 36 patients
We read this dependent shaft label analysis before the plan to establish prior scope. It informed the question and the reuse statement. No code or numerical outputs from it enter our calculation.
A fixed low frequency score ranked seizure onset contacts less well at named contact boundaries in 34 SEEG patients
[Legacy relation: reproduces] The secondary adjacent-contact calculation agrees exactly for all 126 pair identities, both run score differences, and six local group means. Both implementations use the same feature arrays and patients. This is computational agreement, not an independent cohort replication. The present study adds the paired local-versus-all test and other reported checks.
Selection of contacts in named adjacent pairs with different onset labels reduced mean AUC by 0.130 for a fixed low frequency score in an exploratory test of 34 SEEG patients
[Legacy relation: reproduces] Our separate calculation reproduced its secondary adjacent-contact mean of 0.5078 on the same source features and patients. We found this record after our planned results. This is computational reproduction on shared data, not independent patient validation.
Restriction to boundary contacts reduced mean low frequency score AUC from 0.648 to 0.518 in two 120-second SEEG samples from each of 34 patients
[Legacy relation: reproduces] Our boundary concordance of 0.507811 for 126 named boundaries in 34 patients agrees with this study secondary result. The source data are shared, so this is dependent reproduction. Its shaft label permutation test was not repeated. We read its report after our original calculations.