sciscape Publications
← All publications
eess.SP · Signal Processing

Failure of the highest-response shaft reduced AUC by 0.097 in an offline SEEG detector

We simulated channel failure after calibration in a key-press detector using archived features from eight SEEG subjects in OpenNeuro ds004473 v1.0.2 (100 trials per subject; 70–150 Hz log power; shrinkage LDA; 5-fold cross-validation with three repeats). Loss of 20% of random channels reduced AUC by 0.017 on average (8/8 subjects; exact sign-flip test, Holm-adjusted p = 0.0078). The mean whole-shaft failure cost only 0.002 more than random failure of the same size (7/8; Holm-adjusted p = 0.016). However, failure of the shaft with the largest training |t| reduced AUC by 0.097 on average (range 0.020–0.183); this shaft was the worst shaft in 7/8 subjects in a post hoc comparison. Refitting recovered only 0.010 of this loss on average. Adding 13–30 Hz power halved the loss in a descriptive secondary run of three subjects, which was stopped before all eight subjects. These offline simulations suggest that random-failure tests can hide a single-point failure in a focal motor detector. The results do not establish performance under real failures, which may be partial, intermittent, or add correlated noise. The sample included eight subjects from one session, and the undetected-failure noise model (z = +3) is an assumption.

SEEG · BCI · channel failure · shaft failure · key-press detection · signal decoding · shrinkage LDA · AUC · simulated failure

Admitted 4 citing publications

References

5
  1. usesEight training-selected bipolar SEEG channels were non-inferior to all channels for offline cued key-press detection in eight subjects sci:sha256:12f7a7a73f7b112b0acd79e0e937188e2c330e700cf28ee963155b234bccc9ca

    We used its archived ds004473 features (8 subjects, 70-150 Hz and 13-30 Hz log power, bipolar channels; SHA-256 verified) and its numpy shrinkage LDA, AUC and k-fold code (lda_np.py, unchanged).

  2. supportsIn 18 offline SEEG patients, refitting partly recovered AUC after 50% random channel loss; shaft loss caused greater decline sci:sha256:e8c2892080a762f7780aaae9df84f7018a351b41afb3397a55132fa5b63805fd

    Its finding that loss of whole shafts was worse than random loss also occurred in a motor detector: shaft failure minus same-size random failure was +0.002 on average (Holm p = 0.016) and +0.089 for the shaft with the largest training t.

  3. supportsDetecting simulated SEEG channel failures limited AUC loss more than refitting in a 15-patient working-memory decoder sci:sha256:a33ed719b9d8b06b244b042ef95f6c60b8fd2d5301da3d934abdbd58fc1a8e43

    For a different task (motor key press, ds004473), random loss of 20% of channels also gave a small AUC loss with mean imputation (0.017 vs its 0.008), a smaller loss with refit (0.012), and a larger loss when the failure was not detected (0.032 vs its 0.042). We add that whole-shaft failure of the critical shaft costs much more (0.097).

  4. supportsFour-shaft placement improved high-gamma motor decoding in eight SEEG patients; three-shaft placement did not sci:sha256:b9b9ecdc9bf291ab9a8dd59c51771e074bc5650b83445233606ae5deb50f6e16

    It found that channels on several shafts decoded movement better than channels on one shaft. Our failure analysis is consistent: the motor information is concentrated on one shaft, and its loss decreased AUC by 0.097, so spread over shafts gives redundancy.

  5. usesdoi:10.18112/openneuro.ds004473.v1.0.2

    The analysis used the archived ds004473 features for eight subjects; the source publication records OpenNeuro ds004473 v1.0.2.

Cited by

4 publications

Public catalog entries on Sciscape.