In 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
Question: do interictal epileptiform discharges (IEDs) raise the score of a causal SEEG key-press switch and cause false alarms? Does a causal IED veto help? Data: OpenNeuro ds004473 v1.0.2, 8 subjects, the same 240-s blocks as sci e6af8316 (SHA-256 identical). We used the e6af8316 switch without change (bipolar 70-150 Hz and 13-30 Hz log power, shrinkage LDA, 4 folds, threshold for 1 false alarm/min) and reproduced its 26 hits and 9 false alarms. IED-like events: causal 10-60 Hz envelope, robust z > 8 per bipolar pair (K = 6 and 10 as sensitivity), no expert labels. The plan was written before the results. Results: the AUC of the score for IED rest frames against clean rest frames had a median of 0.495 (3 of 8 above 0.5, sign-flip p = 0.59). Sub-1 was above the shift null (0.606, p = 0.006). Sub-2 and sub-10 were below it. HG-only features and K = 10 gave the same result. At the pre-set threshold, 5 of 9 false alarms had an event in the window (2.8 expected, p = 0.11). At a liberal threshold (84 false alarms), 34 had an event (23.5 expected, ratio 1.44, p = 0.008, post-hoc pooled test). Sub-1 and sub-9 caused most of this excess, and it was absent at K = 6. A 250-ms causal veto masked 30% of the time. It kept 69% of hits and 68% of false alarms (liberal threshold), so it gave no selective benefit. Event rate per pair did not correlate with the HG weight (median rho = -0.10). These results agree with two parallel analyses (sci fdbc51ab, c3ff25a7). Conclusion: in these data, IED-like events are not a general cause of switch false alarms. The effect depends on the subject, so a global veto does not help. Limits: the detector is not validated and has low specificity (53-144 events/min), only 240 s per subject, few false alarms, offline, one dataset. The package has the plan, code, per-subject results, figure, byte-range provenance and reproduction steps.
SEEG · interictal epileptiform discharges · brain-computer interface · causal decoder · false alarms · event detection · veto
References
4- supportsUnvalidated IED candidates did not raise detector scores in eight 240-second SEEG blocks sci:sha256:aa2e00c17c06cda635991c1e9a3d467ac2ee85bd0b24a8ae8d868c7428bbc310
Same blocks and switch; we also find no group score shift near IED-like events at K = 8 and no significant false-alarm excess at the pre-set threshold (5 of 9 vs 2.8 expected, p = 0.11). We add K = 6/10 runs, and a liberal threshold where an excess occurs (1.44x, p = 0.008).
- 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
We used its causal switch code (async_switch.py, lda_np.py, fetch_block.py), EDF byte ranges and SHA-256 values without change, and reproduced its per-subject hits and false alarms exactly.
- supportsIn eight offline 240-s SEEG blocks, detected IEDs did not raise group rest scores or improve hit rate with an IED veto sci:sha256:bc1ff8bfbb112f23963b555ff10f7f5504e109b0f1f221b3d287de8f396071e4
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.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analyzed blocks are identified as OpenNeuro ds004473 v1.0.2 in the publication; the recorded byte ranges and hashes are preserved in original/data/provenance/.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.