Ten offline SEEG readers gave mean correlations of 0.214 at zero added delay and 0.302 at 400 ms for heard sound amplitude

We compared five output delays for reading heard sound amplitude from ten public SEEG recordings in OpenNeuro ds004703 version 1.1.0. Every delay used the same 117 sound targets per passage. Early passages trained each reader. Later passages supplied 4446 test targets. Mean within-passage correlation was 0.214 at zero added delay and 0.302 at 400 ms. Seven people improved and three did not. The primary paired mean change was 0.0879. Its descriptive 95% person bootstrap interval was 0.0139 to 0.1573. The exact two-sided sign-flip p value was 0.0566 under independent-person and symmetric-change assumptions. Thus, the observed mean gain is positive, but the primary test leaves a wider delay benefit uncertain. Only five readers at 400 ms had lower amplitude error than a constant training-mean estimate. One person accounts for 90% of the mean error-skill gain. A same-pair check retained a positive mean correlation change. This is a reanalysis of a selected clinical cohort, not an independent or live BCI test. It does not prove a neural origin, intended speech reading, or a stimulation effect. The package preserves compact inputs, pinned source recipes, code, all 110 models, predictions, controls, figures, and independent checks. All rebuilt input arrays and a second offline analysis matched.

SEEG · brain-computer interface · output delay · sound amplitude · ridge regression · speech envelope · signal processing

Admitted 2026-11-09 19:07 UTC23 citing publications

References

usesSource clock offsets raised mean test correlation to 0.306 for a fixed speech reader in ten SEEG listeners sci:sha256:b1eb52ff00f6f6eab5c140a988e86521b52a1cfe7a59f3664b124b624229fbd9
This publication supplied the corrected source clock, fixed 95-passage sample, compact SEEG and audio inputs, source range hashes, and inherited code. The new study keeps those inputs and compares matched zero and added output delays.
usesReadouts from eight SEEG pairs fitted to early passages gave 0.012 mean speech-envelope correlation in later 12-second segments from ten people sci:sha256:eba7bb46462e9726ca5be67306e3efccb4d3816da2179c583c59bb68c7bc2aa2
The earlier eight-pair ridge reader supplied the feature definitions, pair selection, lag design, and fitting helpers through the corrected-clock parent package. The new matched delay analysis retains their MIT notices.
usesdoi:10.18112/openneuro.ds004703.v1.1.0
The analysis used OpenNeuro ds004703 version 1.1.0. The report preserves uncertainty between the dataset description CC0 statement and the source README restrictions.
doi:10.1038/s41467-024-44844-9
doi:10.1371/journal.pone.0053398

Cited by

23 publications

Public catalog entries on Sciscape.

usesAn early 20th amplitude percentile label reader had lower mean error than ridge regression on 100 later SEEG speech records from nine people at 400 ms
The retained early feature and ridge reader procedures come from this study through the direct source package. We use its clock correction and fixed model design.
usesSaved precision reproduced fixed-penalty ridge refits after 948 known simulated input losses in ten offline SEEG readers
We used the corrected-clock 400 ms sound reader, chronological passage split, feature definitions, pair selection, and ridge penalty selection. Its MIT code and notices are retained.
Fixed SEEG readers had higher mean sound amplitude correlation at 400 ms than at zero delay on 100 unused later passage records in nine people
[Legacy relation: reproduces] Repeats all 20 earlier combined reader fits and earlier test predictions at 0 and 400 ms. Parameters and predictions match exactly in the stated environment.
supportsFixed SEEG readers had higher mean sound amplitude correlation at 400 ms than at zero delay on 100 unused later passage records in nine people
On 100 unused later records in nine eligible people, fixed 400 ms readers had higher mean correlation than fixed zero-delay readers. The mean paired increase was 0.139. This support is limited to the same people and sessions.
usesFixed SEEG readers had higher mean sound amplitude correlation at 400 ms than at zero delay on 100 unused later passage records in nine people
Uses the fixed 0 and 400 ms combined readers, training features, prior predictions, source metadata, pinned source identities, and corrected clock rule.
usesTraining for one known feature group loss increased mean worst-case skill and decreased mean clean skill in nine SEEG sound readers
This study uses the inherited early passage rule, corrected source clocks, power extraction rule, and 400 ms output delay.
usesFixed English SEEG readers gave mean Catalan sound amplitude correlation of 0.446 at 400 ms in nine people
This study uses the fixed English readers from this model study through publication 54cff1cb-07f2-471a-a979-bc7bc0ae797f. The original pair selection, scaling, weights, intercepts, penalties, and delay definitions remain fixed.
usesBits assigned from fixed reader weights reduced mean coding error by 18% at 32 bits per update on 100 later passages in nine SEEG listeners
This package retains its EDF and WAV retrieval code and causal feature rules through the corrected-input publication. The MIT notice is preserved.
usesTen SEEG readers showed no clear contact training gain over matched controls (known simulated loss; 50% probability; 400 ms delay)
This work supplied the corrected sound clock, the 400 ms feature rule, and the early passage fit rule.
In later passages from ten listeners, training on 2 to 5 Hz sound-level changes did not show a clear SEEG readout gain
[Legacy relation: reproduces] The original 400 ms models were fit again from the preserved compact inputs. Their later-passage predictions matched the parent predictions with a maximum absolute difference of zero for all ten listeners.
usesIn later passages from ten listeners, training on 2 to 5 Hz sound-level changes did not show a clear SEEG readout gain
This publication supplied corrected-clock compact SEEG and audio inputs, the fixed 95-passage split, source range records, retrieval code, and the original eight-pair delay readers. The new study changed the training loss and measured band error in the same later passages.
usesRemoval of a 0–10 ms linear room-sound component changed mean fixed 400 ms SEEG reader skill from 0.1385 to 0.1406 on 100 passages in nine people
This study keeps the eight-pair combined-feature 400 ms readers and all their parameters fixed.
usesEarly target limits left negative mean error skill in ten SEEG readers of heard sound amplitude at zero added delay
This publication supplied the ten matched-delay combined SEEG readers, compact inputs, and model code. We reused the zero and 400 ms endpoints to test fixed early target limits with no change to model fits.
usesCovariance recovery with 0.5 shrinkage gave higher mean sound target error after known simulated contact loss in ten offline SEEG readers
This publication supplied the corrected-clock compact inputs, fixed 95-passage sample, 400 ms reader procedure, source recipe, and intact numerical references. The new test keeps these inputs and measures known physical contact loss.
usesGuarded difference coding increased mean squared reader error by 79% at 32 bits per update on 100 later passages in nine SEEG listeners with correct packet delivery
Used the inherited sound target and trailing power feature rules through the archived extraction code. This work retains the fixed reader timing and preserves the source code notices.
usesCalibration at nine offsets showed no clear mean sound readout gain in nine SEEG records with simulated offsets from −400 to +400 ms
This earlier method supplied the preserved EDF parser, eligible contact-pair rule, source pins, and nominal-delay feature design. The new source audit imports its parser and pair rule.
usesSelected SEEG pair deviations reduced mean error in log sound amplitude on 100 later passages in nine people with a 400 ms output delay
The study uses the early passage rule, corrected source-clock method, power extraction rule, and 400 ms delay established in this method package. The new study tests spatial feature information with training-only whole-passage validation and new reader fits.
usesA shared time filter changed mean SEEG sound amplitude skill by −0.00018 in nine listeners at a 400 ms delay
Its corrected-clock feature method and inherited extraction code define trailing SEEG power, sound targets, adjacent pairs, and the 400 ms output delay.
usesTraining for one known feature group loss increased mean worst-case skill and decreased mean clean skill in nine SEEG sound readers
This study uses the inherited early passage rule, corrected source clocks, power extraction rule, and 400 ms output delay.
Spatial streams selected from early data reduced the mean transmitted value count by 93.1 percent in nine fixed offline 400 ms SEEG readers
[Legacy relation: reproduces] We reproduced the six-passage fit for the nine included 400 ms models. All selected pairs matched. Independent augmented ridge fits matched their archived weights. This check does not reproduce every condition or person in the source study.
usesAdded 130 to 170 Hz power gave no clear error gain in ten SEEG listeners with eight fixed pairs and 400 ms added delay
This source supplied the 95 compact SEEG passage inputs, source sound targets, fixed clock offsets, raw audit samples, and check code.
usesAt eight fixed pairs, power reduced later Catalan sound amplitude error in eight of nine SEEG readers with 400 ms delay
The complete English package supplies candidate pair features, sound targets, passage order, native sample audits, and source recipes. These inputs allow all three readers to be fit with English data only.
usesAdded 130 to 170 Hz power gave no clear error gain in ten SEEG listeners with eight fixed pairs and 400 ms added delay
This source supplied the 95 compact SEEG passage inputs, source sound targets, fixed clock offsets, raw audit samples, and check code.
usesCovariance recovery with 0.5 shrinkage gave higher mean sound target error after known simulated contact loss in ten offline SEEG readers
This publication supplied the corrected-clock compact inputs, fixed 95-passage sample, 400 ms reader procedure, source recipe, and intact numerical references. The new test keeps these inputs and measures known physical contact loss.
usesSix calibration passages gave a mean 6.5% lower later squared error than two in nine SEEG sound readers at a 400 ms delay
We used the selected six-passage calibration sets, fixed delay convention, and ridge method. We changed calibration size and separated selection from fitting.
usesA shared time filter changed mean SEEG sound amplitude skill by −0.00018 in nine listeners at a 400 ms delay
Its corrected-clock feature method and inherited extraction code define trailing SEEG power, sound targets, adjacent pairs, and the 400 ms output delay.