Offsets from labels changed pooled gamma AUC with separate patient score ranges and fixed local ranks in 35 SEEG patients
We audited bipolar relative gamma power in 35 SEEG patients from OpenNeuro ds004100, version 1.1.3. The primary set used 3,463 equal-label contact pairs and the first 20 seconds of two interictal runs. Of 823,942 pooled positive-negative comparisons, 97.19 percent crossed patients. Raw pooled AUC was 0.413. Mean patient AUC was 0.475. Offsets based on observed labels gave pooled AUC values of 0.271 and 0.726 while every patient AUC stayed fixed. These values are exact extrema over fully separated patient score ranges. They are mathematical stress tests, not prediction performance estimates. Patient bootstrap intervals did not establish a gain from two label-free transforms on the fixed pair set. Fresh downloads and SciPy Welch calculations checked all 70 runs and 3,585 bipolar scores. The package preserves a reusable audit tool, proof, source package, spectra, metadata, results, bootstrap draws, and reproduction instructions. The findings apply to this short-window score and supplied labels. They do not validate a clinical localization model.
area under the curve · AUC · SEEG · stereoelectroencephalography · patient grouping · cross-patient comparisons · relative gamma power · seizure onset zone · score evaluation · label-based offsets
References
3- usesComparisons on the same shaft raised mean gamma AUC by 0.14 in two 60 s records from 34 SEEG patients under common average reference sci:sha256:d06b23ca6be13ef1a596fcf0ff5d9cbe23a9a551bf4500fd77a035b3228cf994
This report supplied prior scientific context: the AUC comparison-set identity and preservation of local AUC under separate group transforms. It applied them to electrode shafts. The new study applies these facts to patients and adds separated-block extrema and exact patient comparison counts. We did not use its scores or spectra.
- usesFree labels on mixed pairs gave a mean gamma AUC range of 0.098 in 35 HUP SEEG patients sci:sha256:ff53ecdb170807e6aa4016eead57df4b4e0d12ee020ec266fdd8f6e4aad59f47
This publication supplied the fixed 35-patient cohort, bipolar and common-average pair scores, spectra, source metadata, code, and pinned EDF retrieval records. We preserved its complete package and checked its source data. These files enabled the new patient comparison audit.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analysis uses the fixed 35-patient cohort and source signals from OpenNeuro ds004100 version 1.1.3.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesPatient score offsets changed pooled STE HFO AUC from 0.239 to 0.787 with fixed channel ranks in 18 SEEG patients
We extend its separate patient score block audit from gamma power to released HFO counts. Its unchanged audit module confirms our decomposition, ranks, and endpoints. We do not claim a new general theorem.