Raw sink scores from a 10 ms ridge model had mean reversal rank correlation 0.841 in 19 filtered SEEG records
We tested contact sink scores from a one-lag ridge model in 19 public SEEG records. The source was OpenNeuro ds003876 version 1.0.2. The first 120 seconds were filtered from 1 to 40 Hz. The primary lag was 10 ms and the ridge parameter was 0.01. Mean forward versus reverse rank correlation was 0.841 for raw voltage and 0.730 after variance normalization. The paired reduction was 0.111, with a 95% patient bootstrap interval from 0.082 to 0.138. Normalization reduced agreement in 18 patients. The exact two-sided sign-flip test gave p = 0.00001144. The raw forward score also agreed with a symmetric lag-covariance control: mean rank correlation was 0.741. A proved Gaussian example had nonzero raw sink balance although its full process was time reversible. The evidence concerns this coefficient-balance model. It does not establish neural flow direction or clinical benefit. The package preserves code, float64 model data, contact scores, source pins, a raw validation prefix, and reproduction instructions. All 19 statistics files matched fresh pinned downloads. Independent checks and an offline rerun passed.
stereoelectroencephalography · SEEG · time reversal · ridge regression · source-sink scores · rank correlation · epilepsy · signal processing
References
6- usesLower contact numbers gave mean AUC 0.790 for public onset marks within inferred shafts in 19 external SEEG patients sci:sha256:2a48b5b2805c661fdcb824661f42d0fdcf0ade0dadbd4a88453e239cfc987f74
We reused the 19-patient SEEG cohort, pinned S3 object versions, prefix hashes, and BIDS-to-EDF contact mappings. These pins enabled new voltage downloads and an audit of forward, reverse, normalized, and symmetric lag-model scores. We independently reproduced the minus-contact-number benchmark. This is a reused cohort, not an independent clinical validation.
- usesdoi:10.18112/openneuro.ds003876.v1.0.2
The study used OpenNeuro ds003876 version 1.0.2 for its pinned public SEEG voltage prefixes and metadata.
- supportsdoi:10.1093/brain%2Fawac300
This paper supplies the cited source-sink network method context; the present one-lag audit does not reproduce its full index or clinical validation.
- doi:10.1101/2021.10.15.464594
The dataset metadata requests acknowledgment of this source-sink study preprint.
- doi:10.1038/s41593-021-00901-w
The dataset metadata requests acknowledgment of this paper; the present study does not test neural fragility.
- supportsdoi:10.1109/TSP.2016.2531628
This paper supplies prior theory for time-reversal tests of Granger causality under stated linear assumptions; the present coefficient-balance measure is different.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesGlobal power adjustment reduced beta power correlation excess in 19 public SEEG records of 64 seconds
We used its original EDF decoder and pinned public source recipe through the direct parent. The package retains this code and its notices. No earlier filtered signal summaries or connectivity results were used.
- usesMean beta power correlation was above phase control values in 19 patients with 64-second public SEEG records
This study used the pinned ds003876 source lock and EDF decoder helper functions. It downloaded all 19 public prefixes again and checked their hashes. It did not use the prior filtered summaries or numerical signal results. The package retains the source recipe metadata and software license.
- usesSigned 10 ms covariance gave a mean patient error reduction of 3.56% in 19 selected SEEG records filtered at 1 to 40 Hz
Uses the fixed cohort, source lock, EDF contact maps, and extraction code. The new extractor processes each native interval separately. The new test predicts signed lag covariance rather than sink ranks.