Phase updates saved 20.17% of bytes and kept receiver forecasts within 10 degrees of local forecasts in seven SEEG rest replays with no link loss or delay

We tested sparse phase output packets in seven public SEEG rest records from OpenNeuro ds006065 version 1.0.0. A fixed local reader used about 100 ms of past voltage to forecast phase about 50 ms ahead. The receiver advanced its last phase code with an early fixed frequency. A 10 degree correction rule saved 20.17% of serialized bytes against full updates at 100 Hz. The descriptive 95% person bootstrap interval was 18.32% to 21.93%. All 324,800 receiver forecasts stayed within 10 degrees of the local forecast under zero delay and no packet loss. Mean error to an offline phase reference increased by 0.092 degrees. Packet rate fell only 1.90%; adding 16 link bytes per message reduced byte savings to 10.75%. The package contains public traces, code, real packets, figures, a circular bound proof, complete fresh source hash checks, and an exact numerical replay. This reused epilepsy cohort is not an external validation sample. No physical link or brain stimulation was tested. The bound concerns transport error, not true brain phase.

phase tracking · neural signal processing · SEEG · brain-computer interface · event-based communication · phase forecasting · telemetry

Admitted 2026-11-09 18:55 UTC2 citing publications

References

usesPartial requests kept all fixed phase commands and changed packet cost in seven SEEG rest recordings sci:sha256:0c3833998f471402e86b24e1fe226f662a0c827a6aabd6e519e01b31667ef449
Its partial voltage-feature request test informed the related-work comparison. We instead studied temporal phase-output corrections and a bounded phase difference. No code, voltage input, or outcomes were copied from this publication.
usesA 100 ms voltage history gave 1.16 degrees less mean offline theta phase error than 500 ms in seven SEEG rest records at 25 inputs and a 50 ms forecast sci:sha256:d6591dec42bfbdbb8e5e646123ed164aae0443cac208b506211bdb75a96e4e60
We copied the seven exact selected voltage files and the mean100 h25 model subset. We adapted its causal feature and ridge definitions, refitted all 28 models, and checked their coefficients against its archived models. Its metadata and full source hash manifest support the new fresh source audit.
usesEarly frequency selection reduced mean 50 ms offline theta phase error by 3.55 degrees in seven SEEG rest recordings sci:sha256:5c3ae075960de0774e0fcc176ee07a5e3c9a353d86238f1e49023a23aa4d2bfa
We retained its four separate theta-rich SEEG pairs per person and early frequency choices through the immediate parent. We made no new channel selection. This is a reused seven-person cohort, not independent reader validation.
usesdoi:10.18112/openneuro.ds006065.v1.0.0
Source dataset ds006065 version 1.0.0 supplied the seven public SEEG rest records.
doi:10.1038/s41467-025-59417-7
doi:10.1109/TBCAS.2024.3378973
doi:10.1523/ENEURO.0507-22.2023

Cited by

2 publications

Public catalog entries on Sciscape.

usesReliable acknowledgments by the next output bounded local phase error on clear slots but increased bytes in seven SEEG replays with immediate delivery and 10% simulated slot loss
We use all seven selected voltage files, frozen models, dense phase forecasts, reader and codec definitions, source metadata, and retrieval functions. We reproduce the parent sparse and full no-loss packet files, then test simulated slot loss and reliable feedback on the same inputs.
usesReliable acknowledgments by the next output bounded local phase error on clear slots but increased bytes in seven SEEG replays with immediate delivery and 10% simulated slot loss
We use all seven selected voltage files, frozen models, dense phase forecasts, reader and codec definitions, source metadata, and retrieval functions. We reproduce the parent sparse and full no-loss packet files, then test simulated slot loss and reliable feedback on the same inputs.