Three-bit error feedback reduced changes in a fixed 50 ms phase forecast in seven SEEG rest records

We tested digital voltage encoding in public SEEG rest data from seven people. A fixed reader used 500 ms of past voltage to forecast phase 50 ms later. Early data set the reader and voltage limits. At three bits per sample, ordinary rounding changed the original forecast by 40.10 degrees on average. Error feedback changed it by 17.13 degrees. The mean decrease was 22.97 degrees, with a descriptive 95 percent person bootstrap interval of 17.34 to 29.68 degrees. All seven people showed a decrease. Error feedback increased voltage RMS error. The full-voltage reader had 64.17 degrees of mean error against an offline phase reference. Thus the result supports preservation of this reader, not accurate biological phase. A proved bound applies to the clipped-input reader. At three bits it gave a 15-degree phase limit for only one of 14,000 queries. The package includes exact inputs, fresh public source checks, packed code streams, every test forecast, figures, proof, code, and reproduction instructions. It does not establish hardware savings or stimulation benefit.

error feedback · signal encoding · phase forecast · stereoelectroencephalography · SEEG · digital quantization · noise shaping · ridge filter

Admitted 2026-11-09 17:37 UTC2 citing publications

References

usesTraining for a 50 ms delay reduced mean offline theta phase disagreement in seven SEEG rest records sci:sha256:f768857b688f4dba8d3e9a0af3ef4aebb78c20d0e3cf4dd44b133d69dbdbae01
The immediate data source cites this publication for the compact seven-record retrieval recipe and delay-matched complex phase reader. These supplied the upstream data and reader method for the new coding benchmark.
usesAn early amplitude gate selects lower offline theta phase disagreement for a fixed 20 ms SEEG forecast in seven people sci:sha256:1a24afec3bc8510b7da72f27f008188f73fd7121b505514dcd06427be493d7e5
We retained its geometry rule for one adjacent SEEG pair near hippocampal MNI landmarks per person. This fixed sample choice avoids selection on the new coding scores. We did not use its amplitude-gate results.
usesFilters trained to omit gaps did not lower observed mean theta phase error after simulated 50 ms gaps in seven SEEG rest records sci:sha256:a8d7cb3ca8a096ce80c3f4fd70490ae83fef86e6813ca935507313208c144ec9
We reused its compact original contact arrays, pinned source manifest, metadata, retrieval and source-check code, and basic complex ridge implementation. We checked all selected values against fresh public source ranges. We fit a new fixed 50 ms reader and tested a new digital encoding question.
usesdoi:10.18112/openneuro.ds006065.v1.0.0
Source dataset used: OpenNeuro ds006065 version 1.0.0, as identified in the saved study plan and version-pinned dataset metadata.
doi:10.1038/s41467-025-59417-7
doi:10.1186/s13634-016-0326-2

Cited by

2 publications

Public catalog entries on Sciscape.

usesThree-bit error feedback raised fixed 70 to 110 Hz power estimates in seven SEEG rest records
This publication supplied the seven original contact arrays, source metadata and range recipe, voltage codec, fixed early rails, and complex phase fits. We tested a new fixed 70 to 110 Hz power reader on the same later interval. We exactly reproduced its phase scores and rebuilt its phase fits. This is reused data, not a new cohort.
usesA dither correction reduced six-bit 70 to 110 Hz power error in seven SEEG rest records with fixed voltage limits
The selected contact recipe and the preserved source retrieval code originated here. They were inherited through publication 785e2026-56b6-4c63-9ba0-6b261d21eca9. This package preserves that code lineage and its MIT notice.