sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.AP · Applications

Site scores reduced alpha-beta role error at omitted links in SEEG and MEG; EEG gains did not pass a fixed reference test

Can one score per recorded site predict alpha-beta coupling roles at omitted links? We tested a fixed graph model on public resting human SEEG and MEG derivatives. Twelve SEEG patients formed the development set. The test reserved 47 other patients and all 19 MEG people, with 27 MEG records combined within person. Each link target was the difference between its two phase-amplitude coupling (PAC) frequency roles. A local-coupling baseline and five link folds controlled model fitting. Adding PAC site scores reduced baseline error by a mean 13.96% in SEEG (pointwise 95% bootstrap interval 11.03 to 17.24) and 8.55% in MEG (5.42 to 11.98). Gains and gains above released shift-reference gains were positive in 46/47 and 19/19 people. The two-modality rule passed after correction for two model claims (Holm p=0.000003815). Reversing phase-phase site scores did not pass. A fixed external test used new phase estimates in 29 healthy EEG controls from another study. All had positive measured gains, but only 19 had positive excess over the computed shift-reference gain. The stricter three-modality rule failed (Holm p=0.136), despite a positive mean excess interval. A low-local-coupling screen also gave uncertain SEEG gains in 24 eligible patients. A post hoc EEG check kept the uniform lag law at each link but changed lag sharing across links. Mean gain on reference graphs changed from 0.1239 with shared target lags to -0.0237 with independent link lags. This did not change the fixed failure. A common-source phase model gave opposite roles with equal local coupling and no links between observed sites. Thus the evidence supports conditional interpolation of measured coupling in electrical and magnetic recordings. It does not establish shared neural causes, later-time prediction, or three-modality reference-controlled generalization. We preserve negative results, plans, sufficient inputs, source hashes, code, figures and numerical audits.

cross-frequency coupling · SEEG · MEG · EEG · phase-amplitude coupling · graph prediction · reference dependence · negative result · human neuroscience

Admitted 1 citing publication

References

13
  1. usesdoi:10.1371/journal.pbio.3000685

    We use the released CFS/PAC definitions, source processing and reported group role question to define a new individual-link prediction test.

  2. usesdoi:10.5061/dryad.0k86k80

    We reuse public CC0 version 5 SEEG and MEG coupling derivatives through a pinned repackage.

  3. usesdoi:10.82901/nemar.nm000389.v1.0.2

    This pinned repackage provides separate human cohorts, participant files and source hashes.

  4. usesdoi:10.1007/s10107-010-0419-x

    Graph gradient ranking is established. We use a penalized site-potential fit and do not claim new ranking mathematics.

  5. doi:10.1016/j.conb.2014.08.002
  6. usesHalf-strength prediction increased mean relative squared error against zero prediction for beta triangle cycles in paired 60-second SEEG files from 36 epilepsy patients sci:sha256:968ad632d8b371bb10d3177c47cb2649c512df8e82853e28c421e8b224159d21

    We build on its distinction between repeatability and successful fixed prediction. No code or data are reused.

  7. usesA fixed alpha harmonic phase predictor gave mean loss reductions of 0.0667 when fit and test shared a block and -0.0039 when they used separate blocks in 88 processed ds004504 EEG records sci:sha256:83d2675ab07b1e14df650681ed84e40ac35200c2b525aee1ffb508ebc23bf4fe

    We use its cut-aware four-second EEG window convention. We estimate new phases and a new endpoint; no Fourier inputs are reused.

  8. usesdoi:10.18112/openneuro.ds004504.v1.0.9

    We use all 29 healthy controls and the public processed EEG derivatives for a fixed external-study test.

  9. usesdoi:10.82901/nemar.on004504

    Version 1.0.0 pins the external EEG files and hashes. The DOI is for the mirror dataset.

  10. usesdoi:10.3390/data8060095

    This is the external EEG source descriptor and processing record.

  11. doi:10.1109/ACCESS.2023.3294618

    The external EEG source requests attribution to its first study.

  12. doi:10.1007/s11571-026-10464-w

    The external EEG source requests attribution to this benchmark review. We do not test diagnosis.

  13. doi:10.1088/1741-2552/ae2293

Cited by

1 publication

Public catalog entries on Sciscape.