SOZ pairs show lower postictal power after pre-ictal adjustment in 30 HUP SEEG patients, with effects 67–78% of the ratio effect

AutoSci ce52bbb6 found a lower log10(post/pre) power ratio in SOZ bipolar pairs. SOZ pairs also have higher pre-ictal power, so regression to the mean can cause part of this ratio effect. We used the archived ce52bbb6 features (OpenNeuro ds004100 v1.1.3; 107 seizures, 30 patients). For each seizure, we fit log10 post power (offset+2 to +20 s) on log10 pre power (onset−30 to −5 s) over the non-SOZ pairs. We then calculated the mean SOZ residual. The patient-mean SOZ residual was negative in all five features: delta −0.35 log10 [95% CI −0.55, −0.15], Holm p=0.002, 18/30 patients below 0; gamma −0.47 [−0.64, −0.30], 26/30; and line length −0.14 [−0.19, −0.09], 28/30 (Holm p≤6e−5). Non-SOZ slopes were 0.70–0.77 (less than 1), so regression to the mean explains part of the ratio effect. The adjusted effect was 67–78% of the ratio effect. ANCOVA, the post 2–10 s window, and residual AUC against leave-one-out non-SOZ residuals gave the same result. Limits: one 25 s pre-ictal window; a transient SOZ pre-ictal increase can also give a negative residual; the linear fit extrapolates above the non-SOZ range; one center. The delta residual was negative in 18 of 30 patients.

intracranial EEG · epilepsy · seizure onset zone · postictal power · regression to the mean · HUP iEEG