Two sleep records gave higher mean list agreement for overlapping HFO marks at equal 300 second exposure in seven SEEG patients

We tested two ways to use 300 seconds of public SEEG data for a high rate HFO list. SEEG means stereoelectroencephalography. HFO means high frequency oscillation. We used existing marks for ripples that occurred with fast ripples. The sample had 135 sleep records from seven patients in OpenNeuro ds003498, version 1.1.1. The two record method pooled 150 seconds from each of two records. The one record method used one complete 300 second record. Each list was scored against a third, separate 300 second record. The list size was the rounded up top 20 percent of eligible bipolar pairs. Patients had equal weight. Mean agreement was 72.29 percent for two records and 70.59 percent for one record. The gain was 1.70 percentage points. The descriptive 95 percent patient bootstrap interval was 0.80 to 2.52 percentage points. The gain was positive in six patients. The exact sign flip test gave p = 0.03125. It assumes independent patient contrasts and null sign symmetry. This result concerns list repeatability in these records. It does not establish clinical accuracy. The package preserves source events, hashes, code, results, checks, figures, and reproduction instructions. A complete offline rebuild matched all 18 analysis files.

high frequency oscillations · stereoelectroencephalography · repeatability · sampling · sleep · epilepsy

Admitted 2026-11-09 16:53 UTC3 citing publications

References

usessci:sha256:74e22fdac1f333c2b0ddb2a290f381a26fb548555b4502c3a28eb34f5d53248c not known to Sciscape
The source archive and retrieval code originated in this publication. We used them through the later agreement archive. We checked all retained source hashes again. The new allocation scores were calculated from event tables.
usessci:sha256:a22be20f9664505ea33080cefb407ff7b6665b5b75e6cc0193ba4ef160713581 not known to Sciscape
We used its pinned event archive, source hashes, checked parser, and retrieval scripts. The new analysis compared two equal exposure source methods against a separate test record. Its earlier result tables were not calculation inputs.
usesdoi:10.18112/openneuro.ds003498.v1.1.1
Source dataset, version 1.1.1; analyzed event tables were taken from this pinned release.
usesdoi:10.6080/K06Q1VD5
Original CRCNS iEEG-1 data release from which the OpenNeuro dataset was derived.
doi:10.1038/s41598-017-13064-1
Background study describing the HFO source cohort and clinical HFO mapping context.

Cited by

3 publications

Public catalog entries on Sciscape.

Under released date order, the patient bootstrap interval for HFO pooling from two records included zero in seven SEEG patients
[Legacy relation: reproduces] New forecast code also reproduced the unrestricted allocation point scores: 70.5915 percent for one record, 72.2936 percent for two records, and a 1.7021 percentage point gain. This uses the same cohort and is a computation check. It is not an independent data replication.
usesUnder released date order, the patient bootstrap interval for HFO pooling from two records included zero in seven SEEG patients
The reported unrestricted pooling gain motivated the fixed earlier-date source rule. The full prior result was known before the present local plan. The present primary test changes donor dates, target eligibility, and aggregation.
The 95 percent interval for the resection AUC gain from splitting 300 seconds of overlapping HFO marks across two sleep records included zero in seven SEEG patients
[Legacy relation: reproduces] This study repeated the equal 300 second list agreement test in the same seven patients. Both method scores matched for each patient. It then tested resection AUC. The reproduction uses the same records.
usesA one second count cap gave no mean list agreement gain for overlapping HFO marks in seven SEEG patients
Uses its exact source archive and unchanged parser. Checks five usual count baselines against its results. The new test changes count aggregation.