Joint selection gave 68.8% later balanced accuracy and lower mean feature correlation at eight contacts in eight SEEG recordings
We tested whether joint selection improved later task state accuracy at eight distinct SEEG contacts. We used OpenNeuro ds004473 version 1.0.2 and inherited features from 2,014 trials in eight people. Both rules used four disjoint bipolar pairs, three power bands, early calibration, and the same logistic reader. Individual pair scores gave 70.00% mean later balanced accuracy. Joint ridge selection gave 68.82%. The paired mean change was -1.18 percentage points. Its descriptive 95% person bootstrap interval was -3.74 to +1.89 points. Mean absolute within-band feature correlation fell from 0.133 to 0.087. Higher mean fit accuracy did not give a mean later accuracy gain. The interval does not establish equivalence. This exploratory result concerns offline event-aligned windows in a reused clinical cohort. It does not test continuous control or brain stimulation. The package preserves code, features, source provenance, models, predictions, figures, checks, and a full retrieval recipe. Fresh raw checks covered one trial in each person.
stereoelectroencephalography · brain-computer interface · feature selection · joint selection · balanced accuracy · feature correlation · later-trial evaluation
References
5- usesContact reuse did not show a mean accuracy gain at an eight-contact limit in later trials from eight SEEG recordings sci:sha256:5521afe57746cb6416438c0681dc8b9a3265fe030839a5954b4f5d947edc95e4
This study uses the exact additive path selector as a secondary reference. The reference reproduces its separate-pair result at eight contacts to the reported precision.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
This study uses the preserved trial features, exclusions, time split, source metadata, and raw retrieval code. Fresh raw checks cover one trial per person.
- usesOne pair per shaft showed no clear gain in lowest balanced accuracy after a known simulated shaft loss in eight later-trial SEEG tests at 16 contacts sci:sha256:5f8c5a234d2daf29596b053b0c7a1a1f7aac5fdbdaf43faeac64da4b037739ad
This study uses the stated limit of additive scores as context. The new comparison tests a joint objective at the same contact limit.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset, version 1.0.2; the package preserves its source metadata, file hashes, and bounded raw retrieval records.
- doi:10.1088/1741-2552%2Facae0a
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesA 32-trial rest covariance update did not establish higher SEEG balanced accuracy than a mean update in eight people
Used its verified feature package, chronological split rule, and greedy additive panel code to construct matched frozen readers with eight contacts.