Heart diseases have been the cause of frequent deaths. It is difficult to diagnose heart problems at every medical centre because of lack of technology and the cost to afford it. This problem has been increasing majorly in rural areas. That is why it is very important to develop an affordable and reliable technology. Artificial Neural Networks (ANNs) is intended towards developing such an intelligent system, which can diagnose whether a patient is suffering from a heart disease or not. The dataset is acquired from the UCI Machine Learning Repository. The training dataset was fed into the network. Error Back Propagation algorithm is the learning network used in the analysis. Artificial Neural Network (ANN) is used to classify and distinguish between absence and presence of disease. The performance measure taken into consideration is accuracy. The targets for the neural network have been classified as 0's (Disease is absent) and 1's (Disease is present). The results obtained from back propagation algorithm using varying number of neurons in hidden layer have been compared in this research work. This system has given the best accuracy (at 80.27%) of diagnosing heart disease when the neurons in the hidden layer are kept at four, with high sensitivity and specificity value. This system provides an efficient application of neural networks for detecting heart diseases.