Interictal SEEG features did not identify high-frequency seizure-onset pairs in 22 HUP patients

We joined two archived feature sets from the same OpenNeuro ds004100 patients: interictal features from 300 s clips and ictal 20-100 Hz onset maps. For each seizure, the target was the top 10% of bipolar pairs by z-scored 20-100 Hz power in the first 10 s after onset. The pre-specified predictor (I3) was the mean within-patient rank of interictal delta power, spike rate and line length, with no fitted parameters. The unit of test was the patient (n=22), using an exact Wilcoxon sign-flip test. I3 median AUC was 0.524 (p=0.35), and I3 did not add information to clinical seizure-onset-zone (SOZ) labels (SOZ+I3 vs SOZ, p=0.92). Within SOZ pairs and non-SOZ pairs, I3 was at chance (0.495 and 0.492). A second interictal clip gave the same result. Clinical SOZ labels gave median AUC 0.555 (p=0.003), but these labels were partly made from the same seizures. Spike rate alone was weakly above chance (0.554, p=0.039, uncorrected). A sham target before onset gave chance performance for all predictors. With the line-length onset map as the secondary target, spike rate and I3 were above chance, and SOZ plus spike rate was better than SOZ alone (post hoc, p=0.008). In this one-center dataset, a short interictal recording did not locate the high-frequency seizure-onset pairs. These results do not support using the tested interictal features in place of recorded seizures to select high-frequency onset targets. The study includes one center, 22 patients, 79 seizures, and one definition of onset; electrode coverage was selected clinically, and only the four stated hypotheses were confirmatory.

stereo-EEG · SEEG · epilepsy · seizure onset · interictal · high-frequency power · line length · seizure-onset zone · AUC · OpenNeuro ds004100