Post-stimulus hard negatives reduce visual-stimulus response in a SEEG movement detector, but movement-vs-rest performance was not preserved in 7 patients
In an offline analysis of archived 70–150 Hz features from seven patients in OpenNeuro ds006253, adding post-stimulus windows to the negative class increased the movement-versus-post-stimulus test AUC by a mean 0.050 (7/7 patients; exact one-sided sign-flip p = 0.0078). Movement-versus-pre-stimulus-rest AUC fell by a mean 0.031; non-inferiority within a 0.02 margin was not shown (5/7 patients above the margin; p = 0.63, Holm-adjusted p = 0.63). Two patients accounted for most of this loss. With the standard decoder, post-stimulus windows were harder negatives than rest windows (mean AUC difference 0.136; 6/7; Holm-adjusted p = 0.047). The result supports a trade-off for this stimulus-related false-positive source, not a general performance gain. This was an offline analysis of one task and dataset, using one feature, one shrinkage LDA, and one time-ordered split; only trials with reaction time at least 0.5 s were included. An online BCI receives continuous data rather than trial windows.
stereo-EEG · brain-computer interface · movement detection · hard-negative calibration · shrinkage LDA · high-gamma features · visual stimulus · OpenNeuro ds006253
References
6- usesLabel-free high-gamma variance did not select SEEG channels that detected movement better than random sci:sha256:b8e1a1f8ce906806aa5c52b1169e9cd151cb002d0db33088aaa3885136a94004
Copy of the ds006253 features with SHA-256 manifest and the numpy shrinkage LDA code.
- usesIn seven SEEG patients, movement-detector timing was stimulus-locked (mean latency slope 0.02); post-stimulus hard negatives did not change it sci:sha256:43b8e57b45f4eb585d0522b1bda62811d03a481ec4938c65c935192852f21873
Showed that first detections follow the stimulus; motivates the hard-negative test.
- supportsA movement detector produced stimulus-linked false alarms in two new SEEG patients, with effects persisting across seven patients sci:sha256:18b6dd07c7fa2fba85f4f0d322571a3ffe58f5678b9d770a85ff1a069c4b319c
The standard decoder again scored post-stimulus windows higher than rest windows (6 of 7 patients).
- usesLeft/right choice decoding was weak and not significant during movement in 7 SEEG patients sci:sha256:bcb8b5c72b2756110340a2fa1572db53fb1917b9f570015743370319b0cb3a7f
Archived ds006253 70-150 Hz window features and S3 provenance.
- usesIn five patients, a 70–150 Hz SEEG movement detector also scored visual stimulus periods highly sci:sha256:15f05b5465a987bcbeecc6cc162a2bce898262ac7004b2e313258f644479e9f8
Showed that the movement detector responds after stimulus onset; this work tests a calibration fix.
- usesdoi:10.18112/openneuro.ds006253.v1.0.3
The analysis used archived author-derived features associated with OpenNeuro ds006253. Per-patient source keys, ETags, and SHA-256 values are retained in the original provenance files; the feature hashes also match the cited archived source manifest.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesTime-locked training did not reduce movement-onset timing error in seven SEEG patients
We use its hard-negative training (pre-stimulus plus post-stimulus negatives) as descriptive model B.
- usesA pooled SEEG decoder matched bin-specific decoders, but its gain required five times more training windows
Showed stimulus-locked false positives of the movement detector. We use this to interpret the above-REST scores in the pre-onset bins.
- usesReaction-time deconvolution links pre-movement SEEG detector scores to movement preparation in 15 patients
Same detector and the post-stimulus false-positive question; our result suggests post-stimulus hard negatives also remove movement preparation.
- In 8 new ds006253 SEEG patients, post-stimulus hard negatives increased movement-vs-stimulus AUC (8/8, +0.057) but decreased movement-vs-rest AUC by 0.027 (shrinkage LDA, offline 0.5-s windows)
[Legacy relation: reproduces] Same pre-registered tests and unchanged code on 8 independent ds006253 patients; primary and S2 replicated, S1 non-inferiority again not shown.
- supportsA rest-trained SEEG decoder times mouse-movement onset better than a shifted null, but often crosses early
That work found that post-stimulus windows give high movement-detector scores; here a rest-trained detector crossed its threshold 0.3 s or more before movement onset in 53% of held-out trials, which is consistent with high pre-movement scores.