Artificial Neural Networks (ANNs) play an important role in the field of medical science in solving health problems and diagnosing diseases both in critical illnesses and in common diseases. Since it is important to diagnose accurately the people' disease condition, therefore for the precisely diagnosing those condition, we must use appropriate methods that to minimize the errors in diagnosis. So, using an appropriate method to diagnose heart disease and to prevent complications of the disease is an important step toward patients' improvement. Therefore, in this paper the presence or the absence of heart disease of the four datasets using Generalized Regression Neural Networks (GRNN) will be discussed. Each of the four datasets contains of 14 features that they are used to diagnose heart disease with GRNN. In this paper, GRNN have been implemented in MATLAB environment. The aim is maximizing the precision of measurement in accurately diagnosing heart disease in the process of training and testing. By comparing the results of each dataset, we found the best accuracy in the training phase that is equal to 100% which belongs to Switzerland and Long Beach VA datasets, and the best accuracy in the testing phase belongs to the Cleveland dataset that is equal to 96.6667%.