Cardiac imaging plays an important role in the diagnosis of heart disease. This disease is at top of the list among all major causes of death worldwide. Prognostication and in-time diagnosis can help in reducing the mortality rate as well as increases the survival rate of patients. Therefore, in this study, cardiac image-based automatic heart disease prediction using a deep neural network is proposed to reduce the death rate due to heart disease. The proposed system consists of four stages namely, pre-processing, feature extraction, feature selection, and classification. Initially, the images are pre-processed using a median filter. After the pre-processing gray-level co-occurrence matrix (GLCM) features are extracted from each image. To reduce the complexity, among the extracted features important features are selected using the lion particle swarm optimization (LPSO) algorithm. Then, the selected features are given to the deep neural network (DNN) classifier to classify an image as normal or heart failure (abnormal). The abnormal consist of three types namely, heart failure without infarction, heart failure with infarction, LV hypertrophy. Finally, the performance of the proposed approach analyzed in terms of different metrics.