Dilated cardiomyopathy (DCM) is a major cause of mortality among patients(pts) with heart failure. A prognostic model is needed to determine whether or not a pt suffering from certain disease will survive or not. Aim of the present work was to study the effective utilization of an artificial intelligence (AI) technique in the medical domain. In this study we attempted using artificial neural network (ANN) and evolutionary computation genetic algorithm GA to predict mortality in this subset of pts to be compared with traditional statistical method namely Cox Proportional Hazard model. Method: The study population consisted of 93 pts admitted at the critical care department of Cairo university (77M, 15F with a mean age of (54.6±10 yrs), 75.3% were ischemic, 21.5% were idiopathic and 3.2% were diabetic and of the 93 pts 30 died (32.2. WE initially considered the following parameters age, sex, disease, duration, NYHA, S3, HR, MBP, NA, CR, Bilirubin, IV, PVCs, EF, LVEDD, LVESD, FS, size of thallium defect as predictor variables and the dependent variable was mortality. Univariate survival analysis for the predictor variables has been performed at 1, 3, and 5 years. Then for the predictors that were significantly associated with the outcome, a Cox stepwise proportional-hazard regression analysis was applied to identify variables which had an independent and significant association with survival. The GA prediction model only used NYHA, S3, EF, IV and PVCs. The GA started on a randomly generated initial population size of 500, ran for 5oo generations, at each generation the fitness was calculated by their ability to discriminate between survivors and nonsurvivors. The population was sorted by the most fit, a number of best fit individuals were selected to be parent for the next generation. Off springs were created via crossover and mutation. We started a single parameter (EF) then we added the rest of the parameters incrementally (NYHA, S3, IVD and PVCs respectively). The addition was guided by the percentage of correctly classified survivors and nonsurvivors. The ANN used the input as GA. Result: using Cox proportional Hazard. The presence of IV delay, with 5 year survival 31.8%and RR 4.5%, Hyponatremia <130 having 5 year survival 35% with RR 1.57%, and EF<40, permitted 45.6% 5 year survival and RR 1.5. According to these predictors pts were classified into four risk groups; (very low risk, low risk, moderate risk and finally high risk groups. Their estimated 5 year survival was 87.5%, 78%, 32% and zero percent respectively. Using GA: EF 20-30%, NYHA, S3, IV delay and PVCs could construct the final prediction model. When respectively added they exhibited incremental improvement in sensitivity and specificity and could correctly classify pts into survivors and nonsurvivors on the learning phase (85%, 83%) and (75% & 80%) for testing phase respectively. Using ANN: the same set of parameters was used with notion that PVCs did not add much to the predictive accuracy of the final model. The ANN permitted 79% accuracy in predicting mortality, 95% in predicting event free survival and 7% error on the testing phase. Conclusion: This study shows the potentials of neural networks for building a prognostic model in DCM. The traditional statistical technique used in this work namely Cox proportional hazard model, could effectively predict the five year survival in pts with DCM.