The mean rank-one/full loss ratio was 1.1022 for 36 SEEG patients in separate file tests at 1–80 Hz and weight 1 with fixed training scales
We tested covariance prediction between pairs of electrode shafts in public SEEG data. Each of 36 patients had two short interictal file excerpts from OpenNeuro ds004100 version 1.1.3. Rank-one transfer kept one shared pattern per shaft pair. Full transfer kept all fitted patterns. The primary test used 1–80 Hz signals, transfer weight 1, and fixed training scales. The mean rank-one/full patient loss ratio was 1.1022. This is a 10.22% increase. The 95% patient bootstrap interval for the increase was 6.68% to 14.11%. Full transfer had lower loss in 33 of 36 patients. The 30–80 Hz check had no clear mean difference. Equivalent changes of basis within each shaft preserved the comparison under a proved cutoff condition. One patient had 99.90% of total raw full-transfer loss. The pooled raw loss ratio therefore gave a different effect size. All 72 input files were rebuilt from public byte ranges and matched exactly. A full analysis repeat from these rebuilt files matched all saved scores. The package preserves data, source hashes, code, scores, a local plan, and reproduction instructions. This is a new analysis of a known cohort. It does not establish separate neural sources or clinical benefit.
stereoelectroencephalography · intracranial EEG · covariance prediction · canonical correlation · rank-one compression · covariance forecasting · basis invariance · reproducibility
References
4- usesFull covariance had lower shape loss than a model with one fitted shared factor in 36 ds004100 SEEG patient means at 1–80 Hz and weight 0.5 after test scale fitting sci:sha256:3cdfebe7e87a178c354d4b2f81550f86c08cf6600ddfe4ab1a9cd673228867f0
The prior complete-covariance one-factor test motivated this distinct pair-specific compression test. The present analysis uses a fixed training scale and a different loss. It does not add an independent patient cohort.
- usesLeading modes captured 9.67 percentage points more scaled 1–80 Hz SEEG variance than a within-shaft control in separate file excerpts from 36 patients sci:sha256:d007e44f45942965fd32b8b5aeadd980f0970f6c452865a0c82b76362f3e5c8b
This publication supplied all 72 lossless contact covariance files, a measured signal example, source metadata, pinned public ranges, and MIT retrieval code. We used them for a new pair-specific canonical covariance prediction test.
- usesTwo bases of the same SEEG contrast space gave different 1–80 Hz correlation participation ratios in 36 patients sci:sha256:901bade82e1972a491fb8b27d87ddf5ae157f439c20fb612d6d24768419473f2
This publication produced the original public SEEG covariance inputs. We obtained those files through publication 846655ac-754f-456a-863b-cfdbed5ef0ee and preserved their source lineage.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The study used selected interictal SEEG records from OpenNeuro ds004100 version 1.1.3, through the preserved covariance inputs and public byte-range reconstruction recipe.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- 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
We used this covariance forecast study to state the scope of the new contribution. It tested pair-specific covariance compression. Our target is the signed normalized covariance difference between two files. Its numeric scores did not enter the new analysis.
- usesAcross files, all other shafts increased mean 1–80 Hz SEEG reconstruction error compared with the nearest shaft in 36 patients
Used its covariance forecast result to distinguish the present voltage reconstruction question from earlier transfer scores.