Ictal energy-ratio rise localizes clinical seizure-onset channels but does not outperform interictal markers or predict surgical outcome

In 37 patients with stereoelectroencephalography (SEEG) from the HUP iEEG dataset (OpenNeuro ds004100 version 1.1.3), we assessed whether ictal energy-ratio markers at clinician-marked seizure onset identify clinical seizure-onset-zone (SOZ) bipolar channels and whether their overlap with treated tissue separates seizure-free from not seizure-free outcomes. We analyzed the first three ictal runs per patient (104 seizures), using 1 s Hann windows stepped by 0.25 s. The mean energy-ratio z-score from 0 to 10 s had mean patient-level SOZ AUC 0.707 (median 0.741; 28/36 patients above 0.5; sign-flip p < 1e-4). The prespecified Epileptogenicity-Index-style (EI) score had mean AUC 0.584 (median 0.537; 25/36 above 0.5; p < 1e-4); its detector found at least one channel in 58% of seizures. In 35 paired patients, the ER rise did not outperform interictal log delta power or spike rate, while EI was lower than interictal delta power, line length, and spike rate. No ictal overlap measure separated seizure-free (20) from not seizure-free (15) patients (all p >= 0.09); the direction for the fraction of treated pairs among EI-marked pairs was opposite to the hypothesis. A post hoc z threshold of 2.5 raised EI AUC to 0.666 but did not produce an outcome effect. Interpretation is limited by circular clinical SOZ labels and onset times, the first three seizures and 20 s post-onset window, no anatomical control, a single center, pooled therapies, and small outcome groups. The full methods, results, limitations, code, and archived outputs are included in the supporting files.

stereoelectroencephalography · epilepsy · seizure onset zone · energy ratio · Epileptogenicity Index · surgical outcome · signal processing