A joint covariance rule with 0.25 shrinkage gave higher mean later Brier error at ten SEEG contacts in eight clinical recordings
Can covariance improve contact selection for a small brain reader? We tested two matched greedy rules on public SEEG data. The source was OpenNeuro ds004473 version 1.0.2. Both rules used ten signal contacts. They learned from early trials and were tested on later trials. Mean Brier error was 0.18545 for the joint rule and 0.18269 for the separate rule. The difference was +0.00276. The paired 95% bootstrap interval was -0.00418 to +0.01012. The joint rule gave higher mean fit separation and lower mean fit correlation. Those changes did not reduce mean later error. The interval permits benefit or harm. The package preserves 2,014 trial features, 120 models, 20,220 test probabilities, four figures, source checks, and all code. This is an offline test of two task states. It does not test a live brain interface or brain stimulation.
stereoelectroencephalography · contact selection · covariance · Brier error · clinical recordings · later-trial test
References
4- usesAn exact contact rule gave higher training scores and higher mean test Brier error in eight SEEG recordings at ten contacts sci:sha256:e54fc72e94a8f20fdec16402dead5f91fcc234591ff2dd38d069d942e98ca8aa
We use its contact graph code and shaft C exclusion. Its additive-score result led to the new matched covariance test.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
We use its unchanged public SEEG features, trial exclusions, source metadata, and retrieval code.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analyzed SEEG features and source metadata came from OpenNeuro ds004473 version 1.0.2.
- doi:10.1088/1741-2552%2Facae0a
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.