Eight local SEEG pairs give 54.8% mean balanced accuracy for cue direction on later trials in eight participants
We tested OpenNeuro ds004473 version 1.0.1. A fixed linear decoder used the first two thirds of trials for training and the final third for testing. Eight local bipolar pairs gave 54.8% mean balanced accuracy. All eligible pairs gave 51.9%. The pre-cue control gave 53.2%. The paired interval for the gain over that control included zero. The conditional permutation p-value for the eight-pair result was 0.0212, under exchangeability within condition and time strata. The target was the cue label in a visual finger-response task. The evidence does not isolate motor intent. The package contains features, source records, code, predictions, and figures.
stereoelectroencephalography · brain-computer interface · cue decoding · balanced accuracy · bipolar pairs · chronological validation
References
2- usesdoi:10.18112/openneuro.ds004473.v1.0.1
The analysis used data from this exact OpenNeuro version, as stated in the publication and recorded source metadata.
- doi:10.1088/1741-2552%2Facae0a
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesEight bipolar signals selected on early trials gave higher mean key-press AUC than random sets on later trials in eight SEEG sessions
This publication provides literature context for the difference between task-state reading and cue-direction reading. Its uncertain gain over its pre-cue control limits the interpretation of successful task-state discrimination. No code, data file, or model from that publication was used in the present analysis.