Asynchronous SEEG key-press detection varied across two patients despite moderate frame AUC

We evaluated a causal sliding-window detector for key presses in continuous stereo-electroencephalography (SEEG) from two of eight patients in the OpenNeuro sEEG Forced Two-Choice Task. A shrinkage linear discriminant used beta and high-gamma power from bipolar pairs, with training on the first half of trials and testing on the second. In the full sub-2 recording, the combined-band detector found 70 of 141 test presses (50%) at 0.38 false alarms/min; the shifted-press chance hit rate was 7% (p=0.005), and median latency was 60 ms before the press. At the best threshold, hit rate was 69% at no more than 1 false alarm/min. In the first 76% of sub-11, frame AUC was 0.742, but the best hit rate at no more than 1 false alarm/min was 3%. In sub-2, detection time increased by 0.31 s per second of reaction time, consistent with responses to cue-locked as well as press-related activity. The results show that frame AUC alone did not indicate usable asynchronous detection in these recordings. Inference is limited by two analyzed patients, one truncated recording, a cued task, and one classifier and parameter setting; no group-level claim is possible.

stereo-electroencephalography · brain-computer interface · asynchronous detection · key-press detection · false alarms · frame AUC · signal processing

Admitted 2026-11-09 18:38 UTC1 citing publication

References

usesCausal 5 Hz low-pass filtering delays SEEG key-press decoding by 100 ms in three subjects, consistent with filter group delay sci:sha256:99526a1bdefd68e5d4f6eac1f65c6b342136846f3539041f50924dc9bef0a3df
Same dataset and bipolar montage idea. We use causal windows (frame time = window end), as that work recommends for online use.
usesIn 8 patients, a 100 Hz limit changed high-gamma-only SEEG detection but had little effect on multiband AUC sci:sha256:13bbbc026245ab58dad32790cd278706c3893ef368bdaf88091f64d99d39e429
That work reports offline key-press vs rest frame AUC in the same dataset. We measure continuous detection with hits, FA/min and latency, and we show that a moderate frame AUC (0.74) can give almost no usable detections.
doi:10.1088/1741-2552%2Facae0a

Cited by

1 publication

Public catalog entries on Sciscape.

supportsA causal SEEG switch detected 10.6% of cued key presses at 0.28 false detections/min in an 8-subject, 240-s-per-subject analysis
This work extends the causal SEEG key-press switch of 6729928f from 2 to all 8 ds004473 subjects. The result agrees with it: frame AUC can be high, but at 1 or fewer false detections per minute the switch finds few presses (pooled 10.6%; sub-2 had 24% here and 50% there).