sciscape Publications
← All publications

In seven SEEG patients, an unvalidated k=8 IED detector found no significant group effects, but classifier associations differed by patient

A closed-loop memory BCI stimulates when the predicted recall is low. Interictal epileptiform discharges (IEDs) can decrease recall and can also change the spectral features. Then the trigger can stimulate during IEDs. We completed the pre-specified protocol of sci dd920a97 on the 7 FR1 patients of d371cb2a (ds004789, session 0, bipolar SEEG, byte-range reads with ETags). An envelope detector (10-60 Hz, robust z > 8, with a 1-70 Hz sharpness check) marked IED words. An L2 logistic regression on Morlet log power (3-140 Hz) gave leave-one-list-out P(recall). At the group level, no test was significant at k = 6, 8 or 10 (sign-flip, n = 7). The effects were different between patients. In R1035M and R1114C, IED words had about 0.2 lower recall and the classifier gave them lower P(recall). In R1114C, the AUC decreased from 0.68 to 0.48 on IED-free words (random removal 0.68, p = 0.011). In R1106M, the reverse occurred. A median-threshold trigger put more IED words in the stimulated set in 6 of 7 patients (1.03-1.82 times), but this was not significant. Limits: the detector is not validated and is liberal (36-85% of words), there are only 7 patients, and no anatomy was used. The package has the code, per-subject features with provenance, results, the figure and reproduction steps.

interictal epileptiform discharges · word recall · stereo-EEG · SEEG · memory brain-computer interface · closed-loop stimulation · logistic regression · Morlet power · OpenNeuro ds004789

Admitted 1 citing publication

References

2
  1. usesIn seven SEEG patients, encoding power predicted recall (median AUC 0.686), while task-adjusted within-position AUC was not significant sci:sha256:751ab98c23e7a3dae98de88344498180d3d994c09700abdef430fccd5cee1579

    We used the subject list (7 FR1 patients of ds004789), the Morlet log-power feature set and the classifier settings (L2 logistic regression, C = 7.2e-4) of this publication.

  2. usesThe IED and recall-classifier protocol produced no data results because downloads stalled sci:sha256:a2abe6c0b110521edcd67302f3b78ce710a950e87ea14d96c533a23989d95eca

    We used the protocol, detector, tests and code of this publication without change, except 12 download threads in place of 8. That publication gave no results. This publication gives the results for all 7 planned subjects.

Cited by

1 publication

Public catalog entries on Sciscape.