IED-like transients did not raise high-gamma switch scores at group level but caused all false detections in one of eight SEEG subjects

We tested whether interictal epileptiform discharge (IED)-like transients raise rest scores or cause false detections in a self-paced SEEG key-press switch, and whether an IED veto helps. We analyzed one 240-s block from each of eight subjects in OpenNeuro ds004473 v1.0.2 using the archived causal LDA switch. The reproduction check matched the earlier result: high-gamma plus beta detected 26 of 246 presses with 9 false detections. A simple 10–60 Hz amplitude-envelope detector, not validated against expert marks, identified IED-like events. For the planned high-gamma test, median subject AUC for rest frames with an IED versus clean rest was 0.475; the group sign-flip test was not significant (p = 0.48). Two subjects exceeded their circular-shift null 95th percentile. In one subject, all 8 false detections at a descriptive, test-tuned threshold of 2 false detections per minute coincided with an IED (1.7 expected); the offline veto removed these false detections and all 3 hits at the pre-set threshold. Pooled at 2 per minute, 21 of 61 false detections coincided with an IED (13.8 expected), and the veto changed false detections from 61 to 40 and hits from 65 to 56. A post-hoc stricter detector gave the same group-level conclusion. The evidence is limited to short blocks and small false-detection counts; the event detector is unvalidated, the 2-per-minute threshold was test-tuned, and the veto used non-causal labels. These results show no group-level increase in rest scores in these eight blocks, while one subject showed IED-associated false detections.

stereo-EEG · interictal epileptiform discharges · brain-computer interface · false detections · high-gamma activity · signal classification