Ridge 0.1 lowered mean transfer correlation of the first fitted mode between contact label groups in public SEEG data from 36 patients at 1 to 80 Hz

The test used public SEEG data from OpenNeuro ds004100 version 1.1.3. It asked if fixed contact-space ridge 0.1 improved transfer of the first fitted correlation mode between contact label groups to the other released file. The test used two files from each of 36 patients, 1 to 80 Hz data, and equal patient weights. Mean test correlation fell from 0.283559 to 0.270169. The change was -0.013389. Its descriptive 95 percent patient bootstrap interval was -0.018255 to -0.008431. Scores fell in 33 patients. An identity penalty changed scores under equivalent contrast coordinates. A transformed contact metric preserved the scores to numerical precision in the tested bases. All fixed secondary settings are reported. Fresh pinned raw ranges rebuilt all 72 covariance files exactly. All 221 metadata files matched a fresh pinned Git archive. An independent solver checked all 101,160 fits, and all 13 numerical output files matched on an offline repeat. The package preserves inputs, code, fitted weights, scores, hashes, figures, and reproduction instructions. The result does not establish an optimal penalty, anatomical connectivity, long-term transfer, or clinical benefit.

SEEG · canonical correlation · ridge regularization · transfer · contact metric · coordinate invariance · electrode shafts

Admitted 2026-11-09 19:19 UTC3 citing publications

References

usesA contact metric preserved 500 ms ridge lag dependence across equivalent contrast bases in 18 SEEG records from nine adults sci:sha256:68aee65b84532e18c2064117d62625b5f87cddf2110f1b60b61dd4d1f72ef716
This work used the contact metric principle from this prior method. It applied the transformed metric to canonical modes between contact label groups and tested transfer to a separate file in a different cohort.
usesZero cross-contact covariance gave mean shaft modularity of 0.903 in 1–80 Hz bipolar absolute correlation models for 36 SEEG patients sci:sha256:f365fd21652627699c341436d59f0f62e54eb790ccef88d7d597a7d9cdf42dd7
This work used the 72 lossless contact covariance files, the direct signal example, metadata, and pinned public retrieval code from this source to fit and check the separate-file transfer test.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
This study used the HUP iEEG Epilepsy Dataset, OpenNeuro ds004100 version 1.1.3, through the archived covariance inputs and checked source metadata.

Cited by

3 publications

Public catalog entries on Sciscape.

usesCommon average reference forced unit canonical correlation across files for fixed complete contact partitions in 36 SEEG patients with full contrast rank
This study of CCA transfer between shaft contrasts informed our file transfer endpoint and the need to state the measured signal space. We tested a complete contact partition and a structural unit mode.
usesLocal symmetric whitening slightly lowered transfer correlation in SEEG files from 36 epilepsy patients at 1 to 80 Hz
Its fixed ridge transfer result in the same 36 patients motivated this local calibration test. The present study evaluates one separate target block and adds target local covariance calibration.
usesThe mode gap gave an uncertain 0.26% error reduction for a fixed linear model in 18 SEEG test patients at 1 to 80 Hz
This publication supplied the covariance inputs, pinned source records, metadata, and unchanged CCA and raw retrieval code. These files made the new gap prediction test and its fresh checks of all 72 source records possible.