Marked time gave 0.31 percentage points lower mean top 20% FRandR list agreement than count in 135 SEEG sleep records from seven equally weighted patients

We tested whether time covered by high-frequency oscillation (HFO) marks gives more repeatable channel lists than event count. We used 23,528 FRandR marks in 135 five-minute SEEG records from seven patients in OpenNeuro ds003498 version 1.1.1. FRandR means fast ripples that occur with ripples. We kept the same channel pairs within each patient. Each list contained the top 20% of pairs, rounded up, with fractional weights for cutoff ties. We compared all different file pairs and gave each patient equal weight. Mean list agreement was 70.28% for marked time and 70.59% for count. The difference was -0.31 percentage points. The 95% patient bootstrap interval was [-3.34, +3.16] percentage points. The exact two-sided patient sign-flip probability was 0.859. These data give no clear evidence that marked time improves repeatability. They do not establish equal performance. A synthetic split of event intervals changed 68 count lists and no marked time lists. This is a property of interval representation, not a test of detector errors. The package contains source marks, checksums, code, results, and a retrieval recipe. The study tests released marks from one detector. It does not test seizure tissue or surgery outcomes.

high-frequency oscillations · SEEG · intracranial EEG · repeatability · event duration · channel ranking · sleep · epilepsy

Admitted 2026-11-09 18:16 UTC2 citing publications

References

usessci:sha256:74e22fdac1f333c2b0ddb2a290f381a26fb548555b4502c3a28eb34f5d53248c not known to Sciscape
Its pinned public SEEG event archive is the original source of the metadata and event tables reused here.
usessci:sha256:a22be20f9664505ea33080cefb407ff7b6665b5b75e6cc0193ba4ef160713581 not known to Sciscape
We reused its source archive, channel maps, list tie rules, and source reader. We reproduced its full-file count agreement baseline before adding duration scores.
usesSpread 60-second samples gave 0.48 percentage points higher mean overlapping HFO list agreement in 135 SEEG sleep records, with equal weights for seven patients sci:sha256:405e297fae9e0c153d7253bdd497f1e879661f2735443f62f739f452f3258939
Its equal-patient count sampling analysis helped motivate this test of another HFO channel score. The cohort is shared.
usesdoi:10.18112/openneuro.ds003498.v1.1.1
The exact OpenNeuro ds003498 version 1.1.1 event marks and metadata used in this analysis.
usesdoi:10.6080/K06Q1VD5
The original Zurich HFO data DOI recorded by the OpenNeuro dataset and source archive.
doi:10.1038/s41598-017-13064-1
Background source study describing the released HFO marks, cohort, and clinical analysis.

Cited by

2 publications

Public catalog entries on Sciscape.

usesUnder released date order, the patient bootstrap interval for HFO pooling from two records included zero in seven SEEG patients
This publication supplied the checked source archive, source_reader.py, restore_sources.py, and the load_records function. The present study reuses source checks, contact mapping, and validity rules. It adds a new test that uses earlier acquisition dates and equal date weights.
usesA 20 ms onset rule left a mean 25.2 percent of shared STE and MNI marked time uncovered in the first hour of 18 SEEG sleep records
Its marked-time comparison and interval subdivision check informed our representation audit. Our test uses two detector lists in a different public SEEG dataset. We reused no source data or code from this publication. We do not claim a new proof of interval union invariance.