Greater Seizure-1 High-Frequency Onset Extent Predicted Better Cross-Seizure Map Transfer in 20 Patients

This study tested whether the spatial extent of a first seizure's onset map predicted how well that map identified onset pairs in later seizures. We analyzed archived stereo-EEG features from 20 patients and 77 seizures in OpenNeuro ds004100 v1.1.3. Extent was the fraction of bipolar pairs with mean baseline-normalized 20–100 Hz power (HF) z ≥ 2 during the first 10 s after the clinical onset mark; transfer was the AUC of the first-seizure map for the top 10% onset pairs of later seizures. The prespecified prediction that focal maps would transfer better was wrong: greater HF extent predicted higher transfer AUC across patients (Spearman ρ = +0.54, two-sided permutation p = 0.015; n = 20). All 10 patients above the median extent had transfer AUC > 0.5 (median 0.81); the three patients below 0.5 had extent ≤ 0.07. Line length showed the same direction but was not significant (ρ = +0.38, p = 0.10). Within patients, extent did not reliably identify a better template (17 patients, p = 0.15), and the top-10% share of positive z mass did not predict transfer. Extent correlated with onset strength (ρ = +0.84); exploratory partial correlations did not confirm that extent was a distinct predictor. The between-patient result is limited to 20 patients from one center, depends on a small number of low-extent, low-AUC patients, and uses assumed run order as seizure order without an independent validation set. A near-empty first-seizure HF map may warn that one seizure is insufficient for calibration.

stereo-EEG · seizure onset · epilepsy · onset map · map transfer · high-frequency power · line length · AUC · responsive stimulation · BCI