In eight offline 240-s SEEG blocks, detected IEDs did not raise group rest scores or improve hit rate with an IED veto

We tested whether interictal epileptiform discharges (IEDs) cause false detections in a causal SEEG key-press switch. We used the archived switch of sci e6af8316 (bipolar high-gamma and beta power, shrinkage LDA, four folds) without change, and reproduced its hit rates exactly. We detected IEDs with an amplitude rule and no expert labels. The primary rule was pre-specified. A strict rule was added post hoc because the primary rule gave too many events (22–157 per minute). For rest frames, the AUC of decoder scores near IEDs versus other rest frames had a mean of 0.50–0.53 for both rules and both feature sets (sign-flip p ≥ 0.70). One subject (sub-1) had higher scores near IEDs (AUC 0.61–0.76, shift p = 0.005). With the strict rule, zero of nine realized false detections of the HG + beta switch followed an IED. A 0.30-s veto after each IED decreased mean hit rate at no more than one false detection per minute (0.242 to 0.138 or 0.198). A sham veto of the same duration gave the same decrease. Thus, for this decoder and these short blocks, IEDs were not a main source of false detections. The package contains the plan, code, per-subject results, score traces, figures, and data provenance.

SEEG · interictal epileptiform discharges · brain-computer interface · high-gamma · beta power · key-press decoder · shrinkage LDA · false detections · causal decoder · IED veto · OpenNeuro

Admitted 2026-11-09 19:17 UTC1 citing publication

References

usesRe-referencing improved high-frequency SEEG detection over the recorded reference, but no montage was best across eight subjects sci:sha256:fde123191b6f52b597384fb536814e8db9f51ec1854910522f07a26b58c04704
Uses its numpy shrinkage LDA code (lda_np.py) through e6af8316.
usesA 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 sci:sha256:bc12d914ad84b0e30b07d04c257741e53711a80db1f9fe906cd0ec29b0929591
Uses its causal switch code, data blocks, events and archived results; reproduces its hit rates exactly.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset used for the eight 240-s SEEG blocks; the recorded analysis specifies ds004473 v1.0.2.

Cited by

1 publication

Public catalog entries on Sciscape.

supportsIn eight SEEG subjects, unvalidated IED-like events did not raise causal switch scores at group level; a global veto removed hits and false alarms similarly
Independent implementation with a causal detector gives the same group null for IED vs rest AUC (median 0.495, p = 0.59), the same sub-1 effect (AUC 0.606, shift p = 0.006), and a non-selective veto.