A 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

One reader was trained on seven public SEEG rest records. It was tested on two records from a separate dataset. No target phase labels entered shared-model training or selection. The reader used early target voltage for pair selection and scale. At a nominal 50 ms delay, its mean absolute error against a centered theta phase reference was 58.38 degrees. Local ridge readers had 58.29 degrees of error. The difference changed direction between records. These means do not establish equal performance. The shared reader was within 45 degrees at 50.02 percent of target test times and 38.59 percent on independent white noise. The centered reference shares past input and uses future voltage. Thus this score does not establish prediction of independent future brain activity. The package keeps nine voltage subsets, fixed reader weights, code, figures, hashes, and reproduction checks. It also records a corrected electrode-name mapping and the repeated analysis. Results describe two new rest records. No stimulation was tested.

SEEG · phase estimation · signal transfer · ridge regression · theta phase · neural signal processing

Admitted 2026-11-09 18:40 UTC16 citing publications

References

usesA 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 sci:sha256:f95e54eb370bc2e0e49af090e06dc6d39029b2a4542255ccea2520190080d0a9
This work uses its seven verified SEEG voltage subsets, source provenance, and MIT phase-reader functions to train one shared source model.
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
This work uses its centered theta reference, past-buffer features, temporal split, and ridge model design, as inherited through the source package.
usesdoi:10.18112/openneuro.ds006065.v1.0.0
The seven source records and selected bipolar-pair arrays are from OpenNeuro ds006065 version 1.0.0.
usesdoi:10.18112/openneuro.ds005670.v1.0.0
The two held-out target records are from OpenNeuro ds005670 version 1.0.0.

Cited by

16 publications

Public catalog entries on Sciscape.

usesA five-bit difference code used fewer bytes and caused less mean phase change than a six-bit sample code in a fixed reader on two depth SEEG rest records
We used all nine selected native-voltage arrays, the early pair selection, the fixed complex reader, source metadata, and source hashes. We checked each clean reader output and reference phase against the archived input estimates.
usesA learned state reduced mean 7 Hz filter phase error after 100 ms simulated gaps in nine SEEG rest records with known initial states
This publication supplied all nine selected voltage extracts, corrected contact mapping, and inherited source provenance. New public prefix downloads reproduced every selected voltage value.
usesTwo exact bounds certified a 7 Hz SEEG filter phase after simulated gaps in nine rest records under voltage caps
The two ds005670 extracts use this study's corrected EDF contact mapping and early pair selection. The new source recipe retains that mapping and tests exact voltage equality.
usesIndependent contact phases gave median gamma network stability above 0.96 in models from two SEEG rest records
We reused its source hash list, retrieval code, and metadata audit for ds005670 version 1.0.0. We retrieved both full records again and checked fresh metadata. We used a new selection based only on labels and metadata.
usesA 0.95 gap map bound limited state sensitivity but increased phase error under dense irregular loss in nine SEEG rest records
Supplied the corrected EDF contact mapping and voltage derivation through the retrieved parent package.
usesIndependent contact models gave unequal bipolar prediction gains at 20 ms with 100 ms histories in two SEEG rest records
The EDF parser and complete-source range downloader were adapted through the parent package. Their MIT terms and source credit are preserved.
usesSpectrum controls gave 50.75 percent mean skill and a causal 7 Hz reader gave 50.41 percent at 50 ms in nine SEEG rest records
We use its two ds005670 input selections, corrected EDF contact mapping, original target metadata, full source checks, and EDF parser.
usesWith shared input history, singular value clipping bounded state sensitivity in 36 fixed 100 ms SEEG readers
We retained the corrected EDF channel mapping and native voltage extracts through the parent. Fresh public-prefix checks matched the selected arrays.
usesA simulated 4 ms delay gave nonzero time reversal contrasts at 20 ms forecasts in independent contact models from two SEEG rest records
The EDF source extractor and full-file range downloader follow its code through the parent package. The package retains its MIT notice and source credit.
usesIndependent phase controls kept positive median bipolar gamma power correlations between shafts in two SEEG rest records
The full-source byte-range downloader came from this work through the cited data package. We preserved the MIT notice and source credit. We used it to retrieve and hash both complete ds005670 records again.
usesIn nine SEEG rest records, a 0.95 gap-map bound limited state growth but raised phase error under dense irregular loss
Supplied the corrected EDF contact mapping and voltage derivation through the retrieved parent package.
usesA product across separate blocks reduced mean gamma score magnitude by over 95 percent for pairs on different shafts in two SEEG rest records
The retrieval and EDF extraction code came from this work through the cited later packages. We preserved its MIT license.
usesA 1 ms constant delay bound kept about 95 percent of mean estimated gamma scores in two SEEG rest records
Used its MIT source retrieval and EDF mapping code through the data parent. Preserved attribution and license terms. Changed the extraction input receipt name to identify fixed expected hashes.
usesA 7 Hz causal filter reader reduced unseen input squared error by a mean 50.4 percent at 50 ms in nine SEEG rest records
This archive supplied the nine native-rate voltage extracts, their hashes, and source timing and contact checks. We used all nine extracts without new pair selection.
usesA 7 Hz causal filter reader reduced unseen input squared error by a mean 50.4 percent at 50 ms in nine SEEG rest records
This archive supplied the nine native-rate voltage extracts, their hashes, and source timing and contact checks. We used all nine extracts without new pair selection.
usesA fixed learned update reduced 7 Hz filter phase error in nine SEEG rest records with simulated 100 ms gaps every 250 ms
The retained public voltage extracts and pair selection use this corrected cross-dataset data package. Its primed-contact name rule for record 01 and the source metadata are preserved.