HFO concentration in 1 s bins gave no clear SOZ rank gain over count on 1,285 channels in the first hour of 18 SEEG records
We tested high frequency oscillation (HFO) timing as a marker of the seizure onset zone (SOZ). We used public stereoelectroencephalography (SEEG) records from OpenNeuro ds005398, version 1.1.3. The main test used released STE detector marks from the first hour on 1,285 selected channels. Each test patient was excluded from model training. Mean patient AUC was 0.7188 for the count model and 0.7104 with a fixed time concentration feature added. The mean change was -0.0084. Its 95% patient bootstrap interval was [-0.0595, +0.0085], with all training fits repeated. This gives no clear gain and does not prove equal performance. A fixed check with at least 20 marks kept 293 channels from 15 patients. Its mean change was +0.0960, with interval [-0.0966, +0.3879]. This check changed the target set and also gave no clear gain. Independent checks passed, and a complete second run reproduced all saved numerical files and arrays. The package contains pinned source bytes, code, results, and reproduction instructions.
high frequency oscillations · seizure onset zone · stereoelectroencephalography · event timing · time concentration · patient bootstrap · channel ranking
References
4- uses20 ms detector matches gave an uncertain SOZ rank gain over combined counts in the first hour of 18 SEEG sleep records sci:sha256:e2fb21fc2b8904e6e332f4adef0382d28035541049ea83d70d8da9f15b9adbb1
This test uses the checked source selection, pinned source files, parser, and retrieval code from this publication. It tests a new concentration feature in the same 18 patients. It is not an independent patient sample.
- usesdoi:10.18112/openneuro.ds005398.v1.1.3
Analyzed source release: OpenNeuro ds005398 version 1.1.3; the saved dataset metadata identifies this exact version and its CC0 license.
- doi:10.1088/1741-2552%2Fac520f
- doi:10.1111/epi.18545
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesIn the first hour of 18 SEEG sleep records, 30-minute STE HFO maps had higher expected top 20% channel agreement than 5-minute maps
Read its whole-hour and half-hour count AUC results before the local plan. These results informed the detector choice. The present duration contrasts and count control use the same public cohort and do not provide independent cohort validation.
- usesMedian HFO duration gave no clear STE SOZ ranking gain beyond count and contact number in 18 SEEG sleep records
This work informed the polynomial count control and the patient bootstrap design. Both models are refitted in each draw, with all copies of each test source patient excluded from training.
- usesMatched HFO counts had lower mean later STE rank agreement than all STE counts in 18 SEEG records
We reused its 94 pinned public source files or EDF headers, manifest, source parser rules, and public retrieval code. We checked all source hashes and refreshed the public byte check. The new analysis uses later count order.
- usesUnrestricted positive SOZ label additions can reverse the mean unfiltered STE count AUC lead over MNI in 18 first-hour SEEG records
This work supplies the pinned public event files, channel metadata, EDF header bytes, source parser, and retrieval code. The present study checks these inputs and uses them to build fixed detector count scores.
- usesEarlier MNI HFO duration gave no clear gain in later duration estimates under patient resampling on 259 SEEG channels from 18 sleep records
We used the exact compressed public source files, version manifest, SEEG selection rules, and retrieval code. These records enabled the source audit and the new duration estimate test.
- usesThe HFO timing AUC change interval included zero within shafts in the first hour of 17 SEEG records under independent, exchangeable patient sampling
Its one second HFO concentration test in the same cohort informed the new timing question. We added local count conditioning, interval joining, same-shaft pairs, contact position control, and fully refitted patient bootstrap fits. We used no result table or code from that study.