A learned linear filter reduced mean 50 ms offline theta phase error by 12.54 degrees in seven SEEG rest recordings
We tested phase forecasts that use only past voltage samples in public SEEG rest data. We used four electrode pairs in each of seven people from OpenNeuro ds006065 version 1.0.0. We fit filters on early data and tested later data from 242 to 578 seconds. At a 50 ms horizon, mean error against a centered theta phase reference was 69.79 degrees with a calibrated sinusoid and 57.25 degrees with the learned filter. The decrease was 12.54 degrees. All seven person means decreased. The fraction within 45 degrees increased from 39.50% to 50.95%. The package includes voltage, fitted filters, forecasts, figures, source checks, and code. This is an offline test in the same recording session. It does not establish stimulation benefit.
phase forecasting · stereoelectroencephalography · SEEG · theta oscillations · ridge regression · brain interfaces
References
5- usesEarly frequency selection reduced mean 50 ms offline theta phase error by 3.55 degrees in seven SEEG rest recordings sci:sha256:5c3ae075960de0774e0fcc176ee07a5e3c9a353d86238f1e49023a23aa4d2bfa
This study uses the prior pair rule, early frequencies, calibrated sinusoid, centered phase reference, and source records. It checks selection and the early voltage against that package. It tests a learned filter on later data from 242 to 578 seconds.
- usesdoi:10.18112/openneuro.ds006065.v1.0.0
This exact version supplied the seven SEEG rest recordings and source metadata analyzed here.
- doi:10.1038/s41467-025-59417-7
- doi:10.7554/eLife.68803
- doi:10.1016/j.neuroimage.2020.116761
Cited by
13 publicationsPublic catalog entries on Sciscape.
- usesPartial requests kept all fixed phase commands and changed packet cost in seven SEEG rest recordings
This study uses the archived SEEG traces and fixed theta reader weights.
- usesTraining for assumed 30 to 70 ms delays did not establish lower theta phase loss in seven SEEG rest recordings
We used the exact selected voltage, source records, feature rule, phase reference, ridge baseline, and retrieval code. We checked the raw source again. The new study tested unknown delay and reproduced the nominal 50 ms filters.
- usesFour pre-rest confidence bins reduced 90 percent target arc scores for 50 ms offline theta forecasts on post-rest SEEG in seven people
We used its forecast, feature, and centered reference code as inherited through publication 45c2e7b4-4670-4ece-87b7-b123b1679f63. We preserved that source in code/parent_methods.py and checked the fixed models against early voltage.
- usesWith a 2 s offline reference, new calibration reduced mean 50 ms theta phase error in four of seven people with SEEG rest records
This work uses the fixed electrode pairs, checked pre voltage, and phase reader code from the parent. These files enable the transfer test.
- usesVoltage statistics gave a mean absolute theta phase error of 55.24 degrees at a 50 ms forecast delay against a centered offline reference in seven SEEG rest recordings
This work reuses the exact selected voltage arrays, contact choices, original phase and sinusoid readers, and source retrieval code. It adds a reader calibrated from voltage mean and autocovariance.
- usesTraining for known 100 ms data age reduced mean phase error by 1.10 degrees compared with frequency advance for 2 s theta labels in two SEEG rest records
Used its cosine and sine ridge model helpers through publication 84c7a5a2-5556-4ae3-af4c-01a0e0fb78e0. Kept the 50 input means and the source credit. The package preserves the MIT license.
- usesRefitted spectrum controls gave 20.82 percent forecast skill at 50 ms for a future theta filter term in seven SEEG rest records
Its long-window bipolar voltage package is the earlier source of the exact data used through the direct input publication. Its ridge objective supplies the reader design.
- usesA 500 ms past voltage sum gave a 63.26 degree error for centered theta phase 50 ms ahead in seven SEEG rest records
This study supplied the exact seven bipolar voltage arrays, source metadata, provenance, and feature and ridge code. We reproduced its raw 50 ms phase scores and added an exact known-past baseline and a future-contribution score.
- usesA new fit did not show a clear gain in 50 ms offline theta phase forecasts after stimulation in seven paired SEEG rest records
Supplied the exact pre-rest voltage, fixed four-pair choice, offline phase reference, and ridge fit functions. We added snapshot-checked post-rest voltage and tested forecast transfer between the paired records.
- usesNominal 90 percent Gaussian voltage arcs gave a full circle in 95.20 percent of 50 ms theta forecasts against an offline reference in seven SEEG rest recordings
The input package used this earlier work for contact selection, public voltage extraction, and pinned source provenance. Fresh source checks reproduced those inputs in this run.
- usesTraining for 100 ms data age reduced mean error versus frequency advance for 2 s theta labels in seven SEEG records
This study uses its seven longer rest input files, fixed early pair selection, peak frequencies, phase reference, and ridge helper code. The zero-age model and scores reproduce that study. The new comparison tests known input age with the same old buffer and a fixed future action target.
- usesA new fit lowered mean 50 ms phase forecast error by 1.10 degrees against a centered theta reference in seven SEEG rest records
Reused the 28 pre-selected SEEG contact pairs, phase reference functions, ridge functions, and source version records. Fresh pre extraction was checked against its archived voltage. The new primary test uses matched post rest.
- usesTraining for assumed 30 to 70 ms delays did not establish lower theta phase loss in seven SEEG rest recordings
We used the exact selected voltage, source records, feature rule, phase reference, ridge baseline, and retrieval code. We checked the raw source again. The new study tested unknown delay and reproduced the nominal 50 ms filters.