Selection for the highest surviving score showed no clear Brier error decrease after known shaft loss in eight later-trial SEEG tests at 16 contacts
We tested whether a rule for the largest training score after any one shaft loss reduces later probability error. We used public OpenNeuro ds004473 v1.0.2 features from eight people. Each rule selected eight pairs from 16 distinct contacts. A logistic reader classified rest and a state before the response. The new rule raised the mean smallest surviving training score by 12.24%. The mean largest later Brier error changed from 0.2191 to 0.2204. The paired change was +0.0014, with a descriptive 95% person bootstrap interval of [-0.0090, +0.0109]. These tests show no clear error decrease. The package includes three figures, all models and predictions, a verified selection method, source hashes, raw range checks in all eight records, and an exact feature rebuild in one record. The data were reused, the windows need event labels, and shaft loss was known and simulated. The result does not establish a live BCI or a stimulation effect.
stereo-EEG · brain-computer interface · shaft loss · Brier error · feature selection · probability decoding · robustness · clinical epilepsy monitoring
References
6- usesEight 16-contact SEEG tests showed no clear key Brier decrease from one pair per shaft under known simulated shaft loss sci:sha256:4049c5a42e64508dd715b6924d74e3805407f51a9c0459e644631fe1e9d33dd6
Use its probability-error question and Brier endpoint for known shaft loss at the same 16-contact budget. This investigation uses rest versus pre-response task state, a fixed reader setting, and a different selection objective. It does not reproduce the earlier recorded-key result.
- usesAn eight-pair rule increased mean surviving high-gamma power after the worst simulated shaft loss in 36 SEEG test records sci:sha256:b5d10a3b7bd8a690c9522d78196853552e773420d35d2fb57e75374318bfbc13
Adapt its total-minus-largest-shaft objective from surviving high-gamma power to early class scores. Add a constraint against shared contacts. Test whether the score gain transfers to later decoder probability error; do not infer decoded information from retained power.
- usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
Use its eight SEEG feature arrays, fixed source trial sample, extraction code, pinned EDF range receipts, and metadata. All 121 copied input files matched its manifest hashes. Fresh source checks passed in every record; a full sub-11 rebuild matched every array.
- One 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
[Legacy relation: reproduces] Reproduce the total-score and one-pair-per-shaft selections at eight pairs and 16 distinct contacts. Both selected panels, intact balanced accuracies, and lowest masked balanced accuracies matched in all eight records. Use these matched controls to test a new surviving-score selection objective and Brier endpoint.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis uses the public OpenNeuro ds004473 release 1.0.2 features and source records.
- doi:10.1088/1741-2552%2Facae0a
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.