A fixed learned model raised the share of commands within 45 degrees of target phase from 39.79 to 51.28 percent in offline 50 ms SEEG replay of seven rest records
A fixed learned phase model gave better command timing than a calibrated oscillator in seven public SEEG rest records. We reused fixed models and tested native-sample replay with a 50 ms delivery delay and a 100 ms minimum command interval. The mean share of accepted commands within 45 degrees of the requested phase rose from 39.79 to 51.28 percent. All seven people showed an increase. The learned model reached 45.47 percent of offline reference crossings within 20 ms, compared with 38.07 percent for the oscillator. A fixed amplitude gate raised phase accuracy to 61.26 percent, but crossing coverage fell to 14.77 percent. A separate Gaussian noise control gave 40.49 percent phase accuracy with a learned model. The centered offline label contains past and future samples. This control shows that accuracy above the uniform-phase rate does not prove prediction of future neural voltage. The study reuses the same sessions as the parent analysis. It does not test stimulation effects or device performance. The package preserves selected public voltage, fixed fits, command indices, results, four figures, source code, and reproduction instructions. Fresh checks verified all seven full source hashes and exact voltage extraction.
stereoelectroencephalography · SEEG · theta phase · phase estimation · closed-loop timing · ridge regression · offline reference · command coverage
References
7- usesA 500 ms past voltage sum gave a 63.26 degree error for centered theta phase 50 ms ahead in seven SEEG rest records sci:sha256:678ac0e1b4e8d00b5e72eb22a8e9bbbfc9c4bd67c224a77925cb4cba677fc26f
We used its analysis of overlap between past voltage and centered phase labels to design and interpret the separate noise command control.
- usesEarly linear fitting reduced mean 50 ms offline theta phase error from 69.52 to 57.75 degrees in seven SEEG rest recordings sci:sha256:6e0d73f3cfa87fcbac3f6dd9697937e67dc539c31c7ddffb19140640d26f1405
We used the seven selected voltage files, 84 fixed 50 ms models, phase reference code, and archived forecasts. These files fixed the input and models for native-sample command replay.
- usesEarly frequency selection reduced mean 50 ms offline theta phase error by 3.55 degrees in seven SEEG rest recordings sci:sha256:5c3ae075960de0774e0fcc176ee07a5e3c9a353d86238f1e49023a23aa4d2bfa
We used the early SEEG pair selection and pinned source retrieval method inherited through the parent publication. They define the public inputs and source checks.
- usesdoi:10.18112/openneuro.ds006065.v1.0.0
Source of the seven rest records and selected SEEG voltage used in the replay.
- doi:10.1038/s41467-025-59417-7
- doi:10.1038/s41598-021-97560-5
- doi:10.7554/eLife.68803
Cited by
8 publicationsPublic catalog entries on Sciscape.
- usesAn early linear model gave 5.20 degrees less error than a calibrated oscillator for a future filter angle at a 50 ms delay in seven SEEG rest records
We used its seven selected voltage files, source metadata, and range hashes. We used its input and oscillator designs. We fitted new models for the future target and all comparisons.
- usesAn early amplitude gate increased prediction skill for the future part of a centered theta label at 50 ms in seven SEEG rest records
Its amplitude-gated phase command replay motivated a joint phase and future error test. We did not use its command outputs or phase model fits.
- usesA shared reader had 58.38 degrees of mean absolute phase error in two new SEEG rest records at 50 ms with a centered theta reference
This work uses its seven verified SEEG voltage subsets, source provenance, and MIT phase-reader functions to train one shared source model.
- usesAn anchor and difference code reduced mean six-bit 70 to 110 Hz power error in seven SEEG rest records with fixed limits
We inspected its BrainVision source format and full-object SHA256 retrieval method. These details guided a new retrieval script. Contact selection and code tests are new. No prior selected arrays, fitted models, or scores enter the result.
- usesSeparate confidence updates for four pulse phases increased mean Brier score in seven later SEEG rest records with a 1 s holdoff
This publication supplied a method for forward phase crossings at native samples. We adapted the method to candidate schedules with fixed holdoff and confidence gates.
- usesA gate based on past voltage raised 4–8 versus 4–10 Hz phase agreement within 45 degrees by 1.73 percentage points in seven SEEG rest records
This study uses its preserved SEEG voltage arrays, early contact selection, fixed phase models, and native command rules. All 224 command streams match exactly. We test new phase band definitions and a past-amplitude gate.
- usesThe mean share of groups of five commands within an offline 45 degree phase limit was 4.35 percent in seven SEEG rest records at a 50 ms delay and a 100 ms minimum gap
We used its seven selected voltage files, fixed phase models, reference functions, and command rules. We computed new complete-group scores and exact count baselines.
- usesForecast vector gates gave higher phase accuracy than amplitude gates after equal early retention in seven SEEG rest replays with an offline reference
This study used the positive phase crossing conditions from the earlier four-phase command replay. The new comparison used one fixed learned forecast and a shared candidate schedule for both score gates.