Objectives: This study presents a head-to-head comparison of the value of cardiac magnetic resonance (CMR)-derived left-ventricular (LV) function and scar burden and positron emission tomography (PET)-derived perfusion and innervation in predicting ventricular arrhythmias (VAs). Background: Improved risk stratification of VA is important to identify patients who should benefit of prophylactic implantable cardioverter-defibrillator (ICD) implantation. Perfusion abnormalities, sympathetic denervation, and scar burden have all been linked to VA, although comparative studies are lacking. Methods: Seventy-four patients with ischemic cardiomyopathy and left-ventricular ejection fraction (LVEF) ≤35%, referred for primary prevention ICD placement were enrolled prospectively. Late gadolinium-enhanced (LGE) CMR was performed to assess LV function and scar characteristics. [ 15O]H 2O and [ 11C]hydroxyephedrine positron emission tomography (PET) were performed to quantify resting and hyperemic myocardial blood flow (MBF), coronary flow reserve (CFR), and sympathetic innervation. During follow-up of 5.4 ± 1.9 years, the occurrence of sustained VA, appropriate ICD therapy, and mortality were evaluated. Results: In total, 20 (26%) patients experienced VA. CMR and PET parameters showed considerable overlap between patients with VA and patients without VA, caused by substantial heterogeneity within groups. Univariable analyses showed that lower LVEF (hazard ratio [HR]: 0.92; p = 0.03), higher left-ventricular end-diastolic volume index (LVEDVi) (HR 1.02; p < 0.01), and larger scar border zone (HR 1.11; p = 0.03) were related to VA. Scar core size, resting MBF, hyperemic MBF, perfusion defect size, innervation defect size, and the innervation-perfusion mismatch were not found to be associated with VA. Conclusions: In patients with ischemic cardiomyopathy, lower LVEF, higher LVEDVi, and larger scar border zone were related to VA. PET-derived perfusion and sympathetic innervation, as well as CMR-derived scar core size were not associated with VA. These results suggest that improved prediction of VA by advanced imaging remains challenging for the individual patient.