A linear model fitted to early rest data reduced mean 50 ms offline theta phase error by 14.64 degrees versus a calibrated oscillator in seven SEEG recordings

We tested whether a linear phase map fitted to early rest data reduces later 50 ms phase error. We used public OpenNeuro ds006065 version 1.0.0, with four selected bipolar SEEG pairs from each of seven participants. Each forecast used 0.5 seconds of past stored voltage. At 50 ms, mean absolute error fell from 69.48 to 54.84 degrees versus a calibrated oscillator. The mean decrease was 14.64 degrees. All seven participant means and all 28 pairs improved. A descriptive participant bootstrap interval was 13.03 to 16.31 degrees. The exact two sided sign flip p value was 0.015625 under its assumptions. Reference filters of 1 and 3 seconds retained the decrease. About 47 percent of linear forecasts still had errors above 45 degrees. The target is an offline centered theta phase reference. It uses future voltage. This same session test does not establish biological phase accuracy, device performance, or a stimulation benefit. All seven complete source EEG files matched the pinned snapshot hashes. Fresh extraction reproduced every input. A full second analysis gave identical results. The package preserves traces, models, forecasts, figures, source checks, code, and instructions.

phase estimation · SEEG · theta oscillations · linear model · signal processing · offline reference · brain stimulation

Admitted 2026-11-09 16:59 UTC2 citing publications

References

usesTraining for a 50 ms delay reduced mean offline theta phase disagreement in seven SEEG rest records sci:sha256:f768857b688f4dba8d3e9a0af3ef4aebb78c20d0e3cf4dd44b133d69dbdbae01
We read its report after our primary analysis to interpret the new comparison. It compares direct delay training with a calibrated phase advance on one pair per participant. Our comparison uses four pairs and a calibrated oscillator.
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 its 28 selected bipolar SEEG traces, source metadata, retrieval code, and weighted oscillator method. This kept pair and frequency selection fixed for the new linear model comparison.
usesdoi:10.18112/openneuro.ds006065.v1.0.0
Source voltage and recording metadata for all seven participants came from OpenNeuro ds006065 version 1.0.0.
doi:10.1038/s41467-025-59417-7
doi:10.1088/1741-2552%2Fab8683

Cited by

2 publications

Public catalog entries on Sciscape.

supportsAn early learned filter reduced mean 50 ms offline theta phase error by 15.73 degrees in seven SEEG rest recordings
We read this report after our primary analysis. Our complex-voltage targets trained in the first block also reduced later 50 ms error against a sinusoid, consistent with its unit-phase model trained in a later early block. Both studies use the same selected traces. This is support across analysis choices, not an independent cohort replication or a comparison between the two learned models.
Eight-bit coefficient storage changed 50 ms SEEG phase forecasts by a mean absolute 1.21 degrees with float64 inputs and arithmetic in seven rest recordings
[Legacy relation: reproduces] After our independent full precision fits and scores, a library search found this matching result. The 50 ms learned filter and calibrated sinusoid mean errors and seven person decreases match within 1e-10 degrees. The same selected signals and people were used. This is a computational reproduction, not an independent cohort. We then fixed and ran a distinct coefficient storage test.