In 35 HUP SEEG patients, ictal high-frequency power rise identified the clinical SOZ (AUC 0.754) but did not separate Engel I from II-IV outcomes (treated AUC difference 0.016)

We tested whether a quantitative ictal-onset marker can explain surgical outcome in SEEG. We used OpenNeuro ds004100 (HUP): 37 SEEG patients and 104 seizures. We read only 60 s around each seizure onset with HTTP range requests. For each bipolar pair, we calculated the rise of 20-100 Hz power in the first 10 s after onset (HFR), and a simplified Epileptogenicity Index (EI). HFR identified the clinical SOZ pairs (mean per-patient AUC 0.754, 95% CI 0.670-0.830, n=35, p=0.00005). EI also identified them (0.704). This positive control is partly circular, because clinicians mark the SOZ from the same seizures. HFR was higher in treated pairs than in untreated pairs (mean AUC 0.70). But this AUC was almost the same in Engel I (0.709, n=19) and Engel II-IV (0.693, n=16) patients (difference 0.016, 95% CI -0.151 to 0.186, permutation p=0.86, Holm p=1.0). The EI version and the top-10% overlap gave the same null result. The fraction of clinical SOZ pairs that were treated did not separate the groups either (0.68 vs 0.60, p=0.50). In an exploratory comparison, ictal HFR was not better than the best interictal feature from publication 6239f6bf (0.751 vs 0.711, p=0.45). Per-patient ictal and interictal AUCs were not correlated. Limits: small groups, ablation and resection mixed, labels only on implanted contacts, simplified EI with a strict threshold. The plan was written before analysis. The package contains the code, per-pair features, the fetch log and the statistics.

stereo-electroencephalography · seizure onset zone · high-frequency power · epileptogenicity index · surgical outcome · permutation test