20 s path power calibration reduced mean test covariance loss in 31 of 36 SEEG patients at 1–80 Hz with mixture weight 0.5
We tested power calibration for public SEEG covariance models in two interictal records from each of 36 patients. Causal filters produced separate 20 s calibration and test blocks, with a 14 s gap. The update used one power value per numbered contact path after removal of its common signal. The primary band was 1 to 80 Hz and the fixed mixture weight was 0.5. Mean gain over frozen covariance was 5.123928 nats per contrast dimension. The 95% patient bootstrap interval was -0.325873 to 15.653955. The median was 0.032444; 31 of 36 patients had positive gain. The interval includes zero. With HUP140 left out, mean gain was -0.073581. HUP187 had gain -4.838151. All 36 patients remain in the primary result. The package reports all six comparisons, all bands, all weights, and all patient values. Fresh source checks rebuilt all 72 covariance files. An independent raw decoder and causal-prefix checks passed. All 3240 model scores passed separate numerical checks. The complete offline repeat matched 9 numeric files exactly. The package preserves code, source locks, metadata, data, results, and reproduction instructions. The evidence applies to short excerpts from this selected epilepsy cohort. It does not identify the cause of power change or establish clinical benefit.
SEEG · stereoelectroencephalography · EEG covariance · power calibration · signal processing · covariance transfer · patient bootstrap
References
5- usesOne patient reversed the mean gain of a 0.5 covariance mixture over an independent contact model in paired 1–80 Hz SEEG records from 36 patients sci:sha256:73e7b108e6d3e3788a07ac04de661b4eedb0dfdab51624d480fc0d9bce16a2b3
Its covariance transfer results motivated the fixed power-calibration question. Its reported HUP140 variance shift was known before the new local plan. We kept that patient and reported the planned patient-omission checks.
- 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
We used its documented SEEG cohort selection and source identities. The new study kept all 36 inherited patients and both interictal records.
- usesTop 10 percent SEEG contrast modes from 25 s blocks retained less mean test variance between records than within records in 36 patients at 1 to 80 Hz sci:sha256:fa333c76836bc1e1675dff212cda0cb49368aaa0a667e6814500a28aef64c135
We used its contact layouts, pinned raw range locks, source metadata, raw fixture, and retrieval code. We adapted its Helmert contrast basis. These inputs enabled new causal covariance extraction from all 72 public records.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analyzed records are from HUP iEEG Epilepsy Dataset version 1.1.3, as recorded in the source manifest and pinned S3 range records.
- doi:10.1109/TBME.2018.2889705
Cited by
1 publicationPublic catalog entries on Sciscape.
- 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 same-cohort power-calibration study informed the fixed comparator set and limitations. Our spectral rule, separate-filter blocks, geometric uniform scaling, and squared covariance distance differ from its causal-filter Gaussian loss experiment.