Sudden cardiac death (SCD) is the final result of cardiac arrest (CA) , defined as an abrupt and unexpected loss of cardiovascular function resulting in circulatory collapse and death. Up to 50% of cardiac deaths in Europe are due to CA. The estimated mortality of CA is approximately 90%, and significant functional and/or cognitive disabilities often persist among those who survive. The advent of the implantable cardioverter-defibrillator (ICD) has revolutionized the prevention of SCD in high-risk patients with reduced left ventricular ejection fraction (LVEF\<35%). However, the algorithm recommended by current guidelines based on LVEF, considered the only parameter to identify high-risk patients, cannot stratify the population and the spectrum of risk with high accuracy. Although the risk of CA is higher among patients with LVEF\<35% and NYHA class\>1, because of the enormity of the population size at risk (i.e., with organic heart disease and LVEF\>35%), most SCD does occur in patients with LVEF\>35%. Additionally, the majority of pts who receive the ICD for primary prevention of SCD will not benefit from the device (in the Sudden Cardiac Death in Heart Failure Trial published in 2005, the rate of appropriate ICD therapy was 21% at five years), and/or will experience some side effects of it. In the Israeli registry of patients who underwent ICD (n= 1729) or cardiac resynchronization therapy (n= 1326), the 12-year cumulative incidence of adverse events was 20% for inappropriate shock, 6% for device-related infection, and 17% for lead failure. Moreover, recent improvements in drug treatment for HF and myocardial revascularization have further reduced the incidence of SCD in pts with low LVEF. Finally, pts with advanced HF are unlikely to benefit from ICD therapy because of the high rates of non-arrhythmic deaths. Therefore, improved risk stratification approaches to guide the selection of pts for ICD implantation are needed, and only a multiparametric approach may aim to personalize the risk prediction of SCD across the broad spectrum of the phenotypes of HF patients. The RESPECT project has been designed to personalize the risk of SCD by integrating and interpreting information highly multidisciplinary: clinical and bio-humoral, genetics and electrocardiography, conventional and advanced cardiac imaging, and data science. The investigators hypothesized that machine learning models capable of dealing with non-linearities and complex interactions among predictors, including genetic, clinical, electrocardiographic, bio-humoral, echocardiographic, cardiac magnetic resonance (CMR), and nuclear cardiology data, would have superior accuracy in predicting the occurrence of SCD compared with the currently recommended metrics of NYHA class and LVEF by two-dimensional echocardiography and that the personalized risk prediction of SCD will translate in more cost-effective use of ICDs. In addition, the investigators will use the multiparametric predictive models to develop a cloud-computing app that will allow clinicians to predict the risk of occurrence of SCD based on specific covariate profiles of individual patients.