The 30–50 Hz aperiodic exponent did not separate SOZ channels in 27 SEEG patients

We tested if simple interictal SEEG features separate seizure-onset-zone (SOZ) bipolar channels from other channels. We used the HUP iEEG dataset (OpenNeuro ds004100 v1.1.3; 37 SEEG patients; 27 patients had enough labeled channels). The pre-specified primary feature was the 30-50 Hz aperiodic exponent. It did not separate SOZ channels in clip 1 (mean per-patient AUC 0.495, p=0.90) or in clip 2 (0.490, p=0.82). Total power (AUC 0.69 and 0.68), line length (0.66 and 0.67) and spike rate (0.64 and 0.64) were higher in SOZ channels in both clips (Holm p<0.02). Line length stayed significant when we compared SOZ channels only with non-SOZ channels on the same electrode shaft (AUC 0.71 in both clips, Holm p<0.004). A leave-one-patient-out logistic model with 11 features had mean AUC 0.71 (clip 1) and 0.68 (clip 2). Spike-rate AUC was higher in seizure-free patients (0.70 vs 0.56, p=0.058; 0.75 vs 0.52, p=0.006), but these outcome tests are exploratory and not corrected. Limits: only the first 60 s of each 5 min clip was used; the spike detector is simple; SOZ labels are clinical; there is no control for brain region or gray/white matter apart from the same-shaft analysis.

stereo-EEG · epilepsy · seizure onset zone · interictal activity · aperiodic exponent · line length · spike rate · spectral power · AUC · OpenNeuro