Automatic spike rate distinguishes SOZ pairs from clinically spiking non-SOZ pairs, but spike-ripples add no separation

In archived per-pair SEEG counts from 28 patients with both seizure-onset-zone (SOZ) and clinically spiking non-SOZ pairs, automatic spike rate distinguished SOZ from spiking pairs (mean per-patient AUC 0.686, 95% CI 0.570–0.793; Holm-adjusted p=0.0067). Spike-ripple rate (AUC 0.694, 0.580–0.799; Holm p=0.0054) and ripple fraction of spikes (AUC 0.645, 0.535–0.752; Holm p=0.016) also exceeded chance. However, spike-ripple rate did not improve separation over all spikes: the AUC difference was +0.008 (95% CI −0.033 to +0.046; Holm p=0.35). Separation was lower than for SOZ versus normal pairs (AUC 0.810). The analysis reused archived detector counts and clinical contact labels from 43 patients across OpenNeuro ds007118, ds007119, and ds007120; 28 patients met the primary eligibility criteria. Limitations include clinical labels based on visual spike review, about 90 seconds of NREM sleep per patient, no visual review of detections, and no outcome-based ground truth.

SEEG · epilepsy · seizure onset zone · interictal spikes · spike-ripples · AUC · OpenNeuro