Mean replacement of strong 2–40 Hz shaft links was lower between equal contact and equal shaft global references than between recordings in 20 s SEEG samples from 36 patients

We compared equal contact weights with equal shaft weights in a global SEEG reference. We used two interictal recordings from each of 36 patients in OpenNeuro ds004100, version 1.1.3. Each recording contributed a 20-second sample. The main graph used the median absolute 2–40 Hz correlation between contacts on different named shafts. We selected the top 10 percent of shaft links. Reference changes replaced a mean 11.93 percent of selected links. Recording changes replaced 34.80 percent. The paired difference was -22.86 percentage points. Its 95 percent patient bootstrap interval was -28.89 to -16.66, under patient independence. All seven fixed checks gave the same direction. The package contains code, source versions, contact covariances, graph scores, patient results, resampling draws, and a raw example. The result applies to these samples and two global references. It does not select a better reference or establish neural interaction or clinical benefit.

stereo-EEG · intracranial EEG · global reference · signal correlation · graph stability · epilepsy

Admitted 2026-11-09 13:31 UTC5 citing publications

References

usesReference changes alter shaft correlation ranks more than time splits in two public resting SEEG records sci:sha256:d8f7fd1d17b47f5f4da5358a99b7fd7b15ee388e10f1df09e419d5b6891a4b70
Its comparison of reference and time changes supplied prior method context. We used a shaft-pair median and a top-link comparison, with different global weight rules, fixed contact sets, and 36 patients from another dataset.
usesPairs that share a contact occur more often in the top 5% of absolute 2–40 Hz bipolar correlations in 20-second SEEG samples from 35 patients sci:sha256:075e07bcc50387a87c25e30bf057a2843e428b5a564e42191f74e4db6a337251
We used its versioned source manifest, source sidecars, range retrieval code, and Fourier filter. We rebuilt all prefixes and checked common run-01 covariance submatrices against its archived files.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
Source recordings and metadata for all 36 patient pairs were taken from this exact OpenNeuro release.
doi:10.1016/j.neuroimage.2018.08.020
doi:10.1093/brain%2Fawad007

Cited by

5 publications

Public catalog entries on Sciscape.

usesA common average reference constraint explained strong negative partial correlation scores under small diagonal loading in 72 SEEG samples
We used all 72 covariance inputs, the fixed 36-patient cohort, source BIDS tables, fixed source object identities, and unchanged extraction code. We rebuilt every input from fresh fixed source byte ranges. The new analysis tests full-order contact precision scores and a finite reference constraint bound.
usesA sample median reference reduced the mean 2–40 Hz power change score after removal of one early high variance contact from the reference pool in 28 second SEEG samples from 36 patients
Its voltage decoder and Fourier filter reached this study through the cited contact exclusion package. The unmodified code hashes are preserved. Its graph results were not inputs to our scores.
usesDiagonal covariance models retained 51% of top shaft links after common average reference in paired 2–40 Hz SEEG samples from 36 patients
This publication supplied the paired contact covariance inputs, source metadata, pinned retrieval records, and extraction code. These inputs enabled the new diagonal covariance benchmark. Its graph result arrays were not used.
usesEqual shaft weights changed repeat agreement of top 10 percent early gamma lists in 5 of 32 SEEG patients
This publication supplied the equal shaft reference rule and its complete shaft duplication property. They were used for the early seizure reference comparison and software control.
usesGlobal reference updates after one shaft exclusion replaced a mean 9.8 to 10.1 percent of strong 2–40 Hz links in 20 s SEEG samples from 36 patients
This publication supplied the 72 covariance inputs, source metadata, retrieval records, raw example, and unchanged extraction code. Those files enabled the new shaft exclusion test. The new study uses the same cohort and is not an independent cohort replication.