Trial resampling reduced offline error at 50 percent acceptance in eight SEEG recordings with 32 calibration trials
We tested error rejection in eight public SEEG recordings from OpenNeuro ds004473 version 1.0.2. Each decoder used 32 early calibration trials and eight bipolar pairs. We kept its class decisions fixed. We ranked later windows by the original probability score or by the 10th percentile support from 256 paired-trial refits. At 50 percent acceptance, the patient mean error decreased from 31.13 to 27.41 percent. The difference was -3.71 percentage points. Its descriptive 95 percent patient bootstrap interval was -5.90 to -1.28 points. The exact two-sided sign-flip p value was 0.0391, under independent-patient and sign-symmetry assumptions. Seven patients had lower error. Ten new seed sets gave mean decreases of 2.44 to 4.02 points. This result applies to balanced, event-aligned windows within one session. It does not establish live BCI control or stimulation benefit.
stereoelectroencephalography · brain-computer interface · selective classification · trial resampling · error rejection · offline decoding
References
4- 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 the exact retained feature arrays, source extraction code, metadata, pinned raw ranges, and raw audit records. All 121 imported files matched its manifest. The calibration and resampling test is new.
- usesEarly confidence scaling decreased mean Brier error from 0.193 to 0.185 in eight offline SEEG patient tests sci:sha256:f3d79f04937a343945e7a53f5b54b00a0284174fca051959a818f1c79d41a7d1
Its probability scaling result motivated a test of error rejection at matched acceptance. Positive temperature scaling preserves ranks. The new trial-refit score can change ranks. We do not reuse its fitted models or code.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis uses the exact feature arrays and source records from OpenNeuro ds004473 version 1.0.2. The saved dataset metadata records this version-specific DOI and CC0 license.
- doi:10.1088/1741-2552%2Facae0a
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesAn eight-pair reader trained from SEEG ensembles reduces mean Brier error compared with half confidence in eight recordings
We use its MIT Newton and scalar EDF check code through the preserved ensemble code. The package keeps the source attribution.
- usesFixed early confidence thresholds reduced mean task state error to 24.5% at 49.6% retention after condition changes in eight offline SEEG tests
This batch rejection study motivated the confidence question. We tested thresholds fixed with separate early data. We did not reuse its resampling score or compare its different reader as a matched baseline.
- usesTrial ensembles reduce SEEG probability error with eight fixed pairs after 32 calibration trials in eight recordings
Newton fitting, independent BFGS fitting, scalar EDF reads, and direct Fourier checks were adapted from this publication. Its trial resampling method informed the paired trial draws. Fitted states and predictions were not reused. The new test averages member probabilities and measures the combined input count.