Radial basis SEEG readers did not improve mean accuracy after condition changes in eight recordings with 32 calibration trials
We tested whether a nonlinear reader improved SEEG task-state readout after a condition change. We used eight people from OpenNeuro ds004473 version 1.0.2. Each reader used eight bipolar pairs and three power bands. Each person supplied 32 early calibration trials. Pair selection, scaling, and setting selection used only these trials. The primary test used 975 later trials from the next condition. Mean balanced accuracy was 63.95% for radial basis readers and 64.02% for linear readers. The paired difference was -0.06 percentage points. The 95% person-bootstrap interval was -1.60 to +1.81 points. These results do not establish equal performance. The package contains source features, raw audit bytes, model states, predictions, code, and figures. The test used event-aligned windows from one clinical recording per person. It does not show continuous command control, transfer to a new day, or a brain stimulation result.
stereoelectroencephalography · brain-computer interfaces · support vector machines · radial basis kernel · balanced accuracy · calibration · condition change
References
4- usesEight fixed SEEG decoders scored 65.8% mean balanced accuracy for response versus rest windows after task condition changes sci:sha256:27c3a162d5f23c03418feb1337bb50c3dbd6fcf78da3fdc3bdc12da973491084
We used its definition of consecutive retained condition segments for later testing. We used 32 calibration trials and compared two SVC kernels. We did not use its fitted readers or outcome scores.
- 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 used the checked power features, retained trial records, source metadata, raw audit ranges, and unchanged retrieval code. These files enabled the new linear and radial basis reader comparison.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The publication analyzes eight recordings from OpenNeuro ds004473 version 1.0.2.
- doi:10.1088/1741-2552%2Facae0a
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesWorst-mask SEEG training reduced mean worst-shaft log loss at C = 1 in eight tests with 32 fit trials
We reuse the checked feature arrays, source records, and 32-trial condition split. The new study fits four linear readers for known shaft-loss cases.
- usesAgreed task labels from separate SEEG shaft groups had 31.0% mean error after condition changes in eight recordings
This publication supplied the feature arrays, trial records, source metadata, raw range recipes, and unchanged retrieval code used in the new analysis.