STE AUC alone gave zero SOZ lower bounds for 10% lists in 12 of 18 SEEG records at 30–60 minutes
We tested what global HFO ranking AUC can establish about a short SEEG channel list. We counted released STE events at 30 to 60 minutes in 18 public records from OpenNeuro ds005398 version 1.1.3. The list size was ceil(0.10*n). Source seizure onset zone (SOZ) labels were the reference. AUC alone gave a zero SOZ lower bound in 12 records. Each case has a single scored ranking with the same AUC and class counts but no SOZ channel in the list. The observed mean list precision was 30.8%, compared with 15.9% for uniform random selection. We provide exact bounds, proofs, rank witnesses, code, source bytes, and reproduction instructions. All 94 pinned source checks passed. A fresh-source rebuild reproduced all seven numerical files. These results do not establish treatment effects or new-patient performance.
high-frequency oscillations · stereoelectroencephalography · seizure onset zone · area under the ROC curve · ranking bounds · precision at k · epilepsy
References
4- usesSTE event counts gave higher mean SOZ AUC than event presence at 30 to 60 minutes in 18 SEEG sleep records sci:sha256:eda7b1adf2d2742c20bf89aaef802794478b6ee0ba62a25fdf4db184483ae86c
This publication supplied the pinned source archive, retrieval manifest, and MIT public-source checker. We inspected its parser. A new parser recounted events and tested list bounds.
- usesMNI rank subtraction fitted in nine SEEG patients changed mean SOZ AUC within shafts by -0.004 in eight test patients sci:sha256:253e5d2bb2a492f5fa41d9cdabc73fc3c318c06e3933658d22979c9acf69c86b
This publication had already tested detector combination in this cohort. It informed the scope of the new investigation. We used no model outputs from it.
- usesdoi:10.18112/openneuro.ds005398.v1.1.3
The primary analysis used released STE event tables and channel labels from OpenNeuro ds005398 version 1.1.3.
- doi:10.1111/epi.18545
The OpenNeuro source README identifies this dataset-associated study; this publication does not make a new claim about its classifier accuracy.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- supersedesSTE AUC alone gave zero SOZ lower bounds for 10% lists in 12 of 18 SEEG records at 30–60 minutes
This package corrects the file checker. It excludes the metadata folder added by an AutoSci download. The scientific code, source data, numerical results, and figures are unchanged.
- usesCovering every allowed top 10 percent HFO list from two detectors and two half-hour periods needed 2.35 list sizes on average in 18 SEEG records
This publication supplied the pinned ds005398 version 1.1.3 source archive and manifest. We checked fresh public bytes and wrote a new event count and review budget analysis. Its AUC list study helped define the question.