Donor trials increased mean Brier error in 11 SEEG patients with four target labels and ridge penalty 1

Can earlier donor trials improve a number reader with four target labels? We tested public OpenNeuro ds005415 version 1.0.0. Donors came from the other three number formats in the same patient. At ridge penalty 1, mean balanced Brier error increased from 0.2811 to 0.3383. The paired increase was 0.0572, with a descriptive 95% patient interval of 0.0162 to 0.0927. Error increased in 10 of 11 patients. The test used 40 target cases and 297 unique later trials. Accuracy did not show a clear gain. The error increase was smaller at penalty 10. Before-onset and shuffled-donor controls did not establish a benefit from correct donor labels. This exploratory result applies to the fixed method, features, and reused clinical cohort. It does not test intended commands or brain stimulation. The package has four figures, compact inputs, all 3,120 fits, predictions, source records, and reproduction code. Fresh retrieval checked all 1,044 input trial ranges. An offline repeat gave exact copies of 88 numerical result files.

stereoelectroencephalography · brain-computer interfaces · number processing · ridge regression · calibration transfer

Admitted 2026-11-09 19:35 UTC0 citing publications

References

usesSEEG readers with eight pairs gave 50.4 percent mean balanced accuracy for number groups on later tests across seen and heard formats in 11 patients sci:sha256:d76aafecfa802c7e76a757e7349784a50a575bf04ecc7c74104880834f906a83
The prior number-group transfer result motivated a test of donor training with four target labels. Its number range, formats, and fitting design differ. It is prior evidence from the reused cohort, not an independent test of this result.
usesOther patients' task records gave correct format and number labels for all 211 later SEEG trials from 11 patients when family and index were known sci:sha256:94b2345085ca7569510dae36a7dc6b9ae85c74707b3663e69d0a09c0f6444bf3
The task-record study showed that known family and index can supply number labels without SEEG. We used that result to limit claims about number content and independent later trials.
usesA fixed SEEG sound versus image reader gave 48.4% mean balanced accuracy across formats on later trials in 11 patients sci:sha256:ff30a8a491e06e1be6ffeac68cd458373b76e5253b9363d45f7828373a56bd26
We reused the final 11-patient features, source receipts, raw audit intervals, event checks, and chronological splits. We changed the label to 1/2 versus 4/5 and compared four target labels with added donor trials. Fresh retrieval checked every reused input trial.
usesKnown task position gave 99.96% format accuracy without SEEG in 13 Numbers patients within source table groups sci:sha256:bc70def735e7ea198640c1e71169cc98139e06582f629a8e32153aa35ebcf330
The task audit showed repeated stimulus orders and a metadata-only label reader. It motivated the stated task-history limit and the control before onset.
usesdoi:10.18112/openneuro.ds005415.v1.0.0
Source dataset used for the SEEG number-group analysis; version 1.0.0 is pinned in the source records.
doi:10.1371/journal.pone.0313155

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.