Debiased squared weighted phase lag index gave mean Dice overlap of 0.286 for top 10% beta graphs across adjacent minutes in 19 SEEG records

We tested repeat agreement of beta SEEG graphs in 19 patients from OpenNeuro ds003876, version 1.0.2. Each graph selected the strongest 10 percent of eligible pairs in a one-minute window. Mean Dice overlap across adjacent minutes was 0.2861 for debiased squared weighted phase lag index. It was 0.3088 for squared imaginary coherency and 0.3114 for the direct squared weighted measure. The planned D-minus-R mean difference was -0.0226. Its 95 percent patient bootstrap interval was -0.0652 to 0.0266. The exact two-sided test over patient sign assignments gave p = 0.3789. This test requires independent patients and symmetric contrasts under the zero null. This sample did not establish a gain in graph agreement. It does not establish equivalence or neural connection accuracy. Earlier results in this sample were known. All 19 records were rebuilt from pinned public signal ranges. All 4,222,800 Fourier values matched the stored values exactly at complex64 precision. Full source precision kept every primary graph unchanged. Independent equations checked all 109,788 eligible pairs. Fresh offline and source replays reproduced all 23 numeric outputs byte for byte. The package preserves code, compact data, provenance, edge scores, fixed checks, figures, and reproduction instructions.

stereoelectroencephalography · phase lag · weighted phase lag index · graph agreement · epilepsy · reproducibility

Admitted 2026-11-09 18:00 UTC5 citing publications

References

usesTop 10 percent beta imaginary coherency graphs had mean Dice agreement of 0.36 across recorded and common average references in 19 SEEG records sci:sha256:cabd47f37ca4f299ef3f9af674f8bb00bcaea94acc9e06355d6f7a282596a680
This publication supplied the coefficient processing preserved through the immediate input publication. We retained its beta bins, segment windows, sample selection, and cross-prefix pair definitions. The present estimator comparison adds new scores to those inputs.
usesPhase variance scaling gave mean Dice agreement of 0.265 compared with 0.309 for unscaled top 10 percent beta graphs in 19 SEEG records sci:sha256:8a5e10f66b2954470b3450027db2a15e4aaf7cfb8713a75299c65e25bd85983e
The earlier phase variance comparison in the same sample informed the present question about weighted phase lag and its bias correction. We read its result before fixing the local plan. We did not use its numeric files.
usesSelected beta imaginary coherency graphs had positive signed scores in the next minute for all 19 SEEG patients sci:sha256:12942fad684c6f2b4b38922ed35443493024d28e55ef71b1f6aaaaf0801646ad
This publication supplied the fixed Fourier inputs, source metadata, range recipes, EDF headers, raw examples, and CC0 helper code. We used them to compare three phase lag measures and rebuilt all 19 records from public source ranges. This study uses the same patients.
usesdoi:10.18112/openneuro.ds003876.v1.0.2
The analysis used the exact OpenNeuro ds003876 version 1.0.2 release, and the saved source metadata records its CC0 license.
doi:10.1016/j.neuroimage.2011.01.055
doi:10.1101/2021.10.15.464594v1

Cited by

5 publications

Public catalog entries on Sciscape.

usesEqual name group weights changed fewer top beta strength nodes than a change of one minute in 19 selected SEEG records
Read its beta graph overlap results before the local plan. They informed the adjacent minute comparison. The present study computes new node strength and selected node endpoints.
usesShaft phase controls preserved bipolar beta power and gave 5.14 percent global rejection under independent uniform shaft phases in 19 SEEG records
This work supplied unchanged Fourier coefficients for 19 records, source metadata, range recipes, and six contact exclusions. EDF and retrieval helpers were adapted from its CC0 code.
usesMean bootstrap Brier gain was 0.0040 for next-minute top 10 percent beta graphs in 19 SEEG records. The interval included zero under patient independence.
Used all 19 archived Fourier inputs, the fixed sample and contact exclusions, public source metadata, EDF headers, range recipes, and CC0 readers. These inputs enabled the new resampling probability benchmark. We independently rebuilt every Fourier value. The patients are the same.
usesA mean 4.56 percent of top 10 percent raw beta PSI edges kept opposite reference signs across two minutes in 19 SEEG records
Supplied the 19 Fourier arrays, metadata-selected sample, six rail-contact exclusions, and pinned source recipes. Its input hashes were checked. This study adds a raw phase slope sign endpoint and an exact model proof.
usesIndependent code reproduced beta shaft weight results in 19 SEEG records; group selection changed the estimated rank agreement gain
This study used its 19 Fourier inputs, raw source examples, source selection records, and provenance for OpenNeuro ds003876 version 1.0.2. This run rebuilt every input from fresh version-specific public source ranges before it completed the calculation audit.