Waiting examples improved mean response separation on later SEEG trials in eight patients
We tested whether waiting examples improve a SEEG response decoder. We used public OpenNeuro dataset ds004473 version 1.0.2 and preserved features from AutoSci publication 109a1ec7-0f28-4374-8c9a-2649077cd6d4. We added new 500-sample waiting windows that ended about 0.25 seconds before the go cue. Waiting and response windows used separate samples. The study included 1790 eligible trials from eight patients. Earlier trials trained two fixed ridge models and set three thresholds. The final 30 percent of source trials supplied 545 eligible test trials. Adding waiting examples changed mean response-versus-waiting AUC from 0.719 to 0.774. The mean paired change was 0.055 (patient bootstrap 95 percent interval 0.022 to 0.093; exact two-sided patient sign-flip p = 0.0234). This result applies to selected event windows within one session per patient. It does not show continuous control or transfer to a new session. The package preserves features, scores, model coefficients, figures, code, raw-data range hashes, and instructions.
stereoelectroencephalography · brain-computer interface · signal decoding · task-negative states · ridge regression · area under the curve
References
2- usesA fixed SEEG decoder separated windows before keypress from rest on later trials in eight patients sci:sha256:bd6c59023a95ca7dc23d6549123fb8e0b3724d75412db49c1e004ce64fc1dd57
We reused the four-band SEEG features, source metadata, and raw retrieval code for ds004473 version 1.0.2. We added separate waiting windows and tested mixed negative training on later trials with separate threshold data.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source SEEG signals, event labels, and contact metadata for ds004473 version 1.0.2.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesFixed SEEG decoders had higher mean early cue call rates than rest call rates on 494 matched later trials in eight people
We read its waiting-state control to define the scope of this study. It tested windows before the go cue in the same cohort. This study tests new windows after the go cue. No features or model scores from that study are used.
- usesEarly cue training gave an uncertain mean gain in minimum AUC on 487 later SEEG trials in seven people
Its test of pre-go waiting examples informed our question. We test post-go cue examples, a minimum AUC across rest and cue states, and fixed threshold policies in the same source cohort. We do not treat the cohorts as independent.
- usesTask background training increased mean detections and background commands in eight later SEEG streams at an early target of one command per eligible minute
Its selected-window study of waiting negatives supplied the question about negative state coverage. We tested continuous task background training with emitted commands. We reused the same public cohort and did not directly repeat its AUC comparison.