SEEG seizure-onset maps overlap resections but do not distinguish seizure-free from failed surgery

We tested whether 20–100 Hz SEEG seizure-onset maps overlap resected or ablated bipolar pairs, and whether overlap is greater after seizure-free than failed surgery. The analysis used 126 seizures from 32 patients in one center (19 seizure-free, 13 failed), from 141 runs among 35 subjects with at least two runs in OpenNeuro ds004100. In seizure-free patients, onset maps ranked resected pairs above other pairs (median AUC 0.69; 15/19 patients above 0.5; p=0.012), while a pre-onset sham map was near chance (median AUC 0.51). The top pair was resected in 28% of seizures, against a 10% chance rate. The primary comparison did not show greater overlap in seizure-free patients: the seizure-free minus failed-surgery mean AUC difference was −0.075 (95% CI −0.20 to +0.05; one-sided p=0.85); onset-map AUC classified outcome with AUC 0.41. Line length gave the same result. Clinical SOZ labels overlapped resections more than onset maps, but that comparison is circular because clinicians used SOZ labels to plan surgery. These findings are limited to one center and small groups; most patients had ablation, and the 10-second onset window is not causal. Resection overlap was a weak external reference for selecting stimulation targets and did not distinguish correct from incorrect maps in this cohort.

SEEG · epilepsy · seizure onset · resection · responsive neurostimulation · high-frequency power · line length · surgical outcome