Training covariance increased mean balanced accuracy by 0.36 percentage points after known simulated loss of one contact in eight offline SEEG decoders
We tested a fixed recovery rule on eight public SEEG recordings from OpenNeuro ds004473, version 1.0.2. Early trials defined all model values. Later trials supplied the scores. The rule estimated missing power inputs from surviving inputs. It removed every selected pair that used the lost physical contact. Mean balanced accuracy increased from 70.00% to 70.37%. The mean gain was 0.36 percentage points. The patient bootstrap interval was +0.07 to +0.65 percentage points. Six patients improved and two declined. The exact two-sided sign-flip value was p = 0.070. Error in the missing score contribution decreased in all patients. This did not ensure better classification. The test was exploratory because prior studies used the same later trials. The package includes a direct readout tool, figures, predictions, source records, and code.
stereoelectroencephalography · SEEG · brain-computer interfaces · contact loss · covariance · balanced accuracy · signal reconstruction · fault tolerance
References
5- usesLoss training raised mean accuracy by 0.34 points under simulated contact loss in eight SEEG decoders sci:sha256:884a92921286ae95392544128af6a4f343937c3b760e03bce164eec4599853d4
This publication supplied the physical contact-loss definitions, mean replacement reference scores, and raw source check code.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
This publication supplied the preserved SEEG feature arrays, trial exclusions, extraction code, source records, and intact model values.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis uses the OpenNeuro ds004473 source recordings, version 1.0.2, as identified in the preserved source metadata.
- doi:10.1038/ncomms13749
- supportsdoi:10.1088/1741-2552%2Facae0a
The cited source study provides the published context for movement-related signals in human SEEG.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesJoint score bounds gave 96.3% mean trial coverage and 34.5% mean output yield in eight later SEEG tests with known simulated loss of one contact
This work supplied the fixed feature and source files, raw extraction code, and covariance readout code used in this study.
- 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 standardized covariance recovery rule and its fixed 0.5 diagonal shrinkage. The new study tests that rule on continuous heard sound targets in a different SEEG dataset.
- usesJoint Gaussian score integration increased mean command log loss by 0.00003 nats per window after one known simulated contact loss in eight offline SEEG readers
This publication supplied the conditional feature recovery rule and fixed 0.5 diagonal covariance shrinkage. We extended it to conditional uncertainty in two command scores.
- 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 standardized covariance recovery rule and its fixed 0.5 diagonal shrinkage. The new study tests that rule on continuous heard sound targets in a different SEEG dataset.
- usesGaussian score integration reduced mean log loss after known simulated shaft loss in eight SEEG recordings
Its covariance conditional-mean reader supplied the comparison method and fixed shrinkage rule. We added probability integration and shaft-loss tests.