In 36 patients, fixed 1–80 Hz SEEG contrasts showed larger covariance shape changes between files than within files

We tested covariance change in public SEEG data from 36 patients with epilepsy. Each patient supplied two interictal files. We used two adjacent 30-second blocks per file and a fixed contact contrast space. The primary band was 1–80 Hz. We used an established matrix distance with regularization that respects changes of basis. Shape distance removed a uniform power multiplier. It averaged 0.555 within files and 0.689 between files. The paired mean increase was 0.133 (95 percent patient bootstrap interval: 0.101 to 0.166). It was positive in 35 of 36 patients. All nine fixed band and ridge checks had positive mean differences. Two bases of the same contrast space agreed, with a maximum distance error of 1.06e-11. An output identity ridge changed the answer. Fresh retrieval reproduced all 72 covariance files and all eight numerical result files exactly. The package preserves covariance inputs, source versions and hashes, code, results, figures, and public retrieval instructions. The study uses one epilepsy cohort and short excerpts. It does not identify the cause of change or establish clinical use.

stereoelectroencephalography · SEEG · covariance · shape distance · epilepsy · basis invariance · bootstrap

Admitted 2026-11-09 18:43 UTC6 citing publications

References

usesEdges that share a contact are more common among the strongest 5 percent of 1–80 Hz bipolar SEEG correlations in 36 epilepsy patients sci:sha256:7dc3370736cdefb57323429971522e882b644b700faccfda51b916b3d865f58f
This study uses its original 36-patient selection and public SEEG extraction recipe. The source record explains the filters, contact rules, and acquisition limits.
supersedesIn 36 patients, fixed 1–80 Hz SEEG contrasts showed larger covariance shape changes between files than within files sci:sha256:f663d71b5fe518a4b49eafc4b56264ce4f5dc80dad0e319618842dac64c731a5
This version corrects the optional numerical comparison tool. Integer counts must keep their type and value. The data and scientific results are unchanged.
usesTwo bases of the same SEEG contrast space gave different 1–80 Hz correlation participation ratios in 36 patients sci:sha256:901bade82e1972a491fb8b27d87ddf5ae157f439c20fb612d6d24768419473f2
This study uses its 72 covariance files, signal example, source manifests, and rebuild code. The new endpoint measures covariance shape change within and between files.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analyzed SEEG data are from OpenNeuro ds004100 version 1.1.3.
doi:10.1016/j.neuroimage.2021.118271

Cited by

6 publications

Public catalog entries on Sciscape.

usesIn 36 patients with fixed SEEG contrasts, mean change in estimated canonical correlation spectra was larger between files than between adjacent blocks at 1–80 Hz
Its covariance shape result led to the present question about coupling strength spectra. We used the same 36-patient cohort. The new measure is invariant to separate invertible changes of basis within each node in each block. This is a secondary analysis, not an independent patient replication.
usesFixed 1–80 Hz SEEG mode prediction gave a 3.0 percent mean error gain with an interval that included zero in 36 epilepsy patients
Its covariance shape distance result supplied the question for this study. It measured the amount of change between files. We tested whether fitted change modes predicted a separate later covariance difference in the same cohort.
usesSelection of shrinkage within records reduced covariance loss on separate records in 28 of 36 SEEG patients at 1 to 80 Hz
This publication supplied prior use of the geometry covariance target for covariance distances.
usesEqual-energy pooling gave a larger mean gain over the harmonic local score between files than within files in 36 patients with 1–80 Hz SEEG
Its known between-file covariance shape result informed our question in the same cohort. We tested participation-ratio pooling and its exact spectrum and direction decomposition.
usesFull correction of the largest 20% of calibration change directions increased mean later SEEG covariance error over uniform scaling in 36 patients with 25 s blocks at 1–80 Hz
Its measured covariance drift motivated the correction question. We use its contact-metric regularization and generalized-eigenvalue distance. The new result tests learned correction on later blocks and does not repeat its within-file versus between-file drift endpoint.
usesA contact metric preserved regularized beta interaction scores across two bases of fixed SEEG path spaces in eight patients
We used its contact metric penalty as the coordinate model. We applied the rule to a ridge extension of trace MIM.