Moving-minute calibration left more than three background commands per eligible minute in seven of eight later offline SEEG task tests
We tested whether calibration limits stop later command bursts in continuous SEEG reading. We used fixed readers for eight people in OpenNeuro ds004473 version 1.0.2. Both rules met an average calibration limit of three background commands per eligible minute. The added rule also bounded each moving minute. The eligible clock excluded windows near recorded keypresses. Seven of eight later offline tests exceeded the moving-minute bound. Background commands fell from 103 to 85, while keypress detections fell from 61 to 55 out of 340. The mean fraction of moving intervals above the bound fell from 53.34 to 42.41 percent. The paired sign-flip reference was p = 0.125. One calibration block had 54 eligible seconds and used a whole-block bound. The package includes exact count code, frozen readers, features, range hashes, figures, and repeat instructions. A fresh source audit checked 16 ranges. Background commands are a proxy for unwanted output. This reused sample did not test live control or stimulation.
SEEG · brain-computer interfaces · command calibration · moving-window limits · background commands · offline evaluation · intracranial electrophysiology
References
4- usesScore reset removed one of 264 commands in eight offline SEEG tests with fixed thresholds and one-second waits sci:sha256:60e2092c111dcd21fb5cec4026e7b1f93b3f4a49c6463d91cd9e9b04fb1d4563
We reused its archived continuous features, original fit inputs, frozen background readers, average thresholds, and source records. We reproduced its level-scheduler baseline of 264 commands and 61 response hits at the target of three background commands per eligible minute.
- usesTask background training increased mean detections and background commands in eight later SEEG streams at an early target of one command per eligible minute sci:sha256:d193eb1dd45105cd24c75e54cf8edf1698b27eda17d7a057b9662c28f33a06be
We used its fixed background reader, average count calibration rule, voltage-disjoint blocks, and preserved fit code. We refit the reader from its original fit inputs and checked parameter errors below 1e-12.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The publication identifies OpenNeuro ds004473 version 1.0.2 as the source of the task data; the cached version-specific dataset metadata records this DOI and CC0 license.
- doi:10.1088/1741-2552%2Facae0a
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesFixed models with past SEEG power gave no clear Brier score gain over cue inputs for 100 ms keypress forecasts in eight later task segments
This package supplied the eight continuous power records, original trial event arrays, contact selection files, source metadata, and pinned EDF range recipes. The new raw check adapts its EDF parsing and independent power calculation. All 72 source receipts matched the archive manifest.
- usesPast score medians did not lower mean background rate error in eight reused SEEG tests with 60 s histories
We used its eight continuous SEEG feature streams, event samples, fixed reader parameters, and source records. We retained its whole-block calibration rule as the comparator. We reproduced all fixed comparator counts before testing the new causal score correction.