Heart Diseases Image Classification Based on Convolutional Neural Network

Currently, deep learning, especially convolutional neural network (CNN), has been widely applied in imaging and computer vision application due to the good accuracy in image classification and pattern recognition tasks. In this paper, a simple CNN model is built for heart diseases image classification. This simple CNN model is directly trained by the heart diseases images without using transfer learning of the pre-trained networks. The pre-trained networks such as AlexNet and GoogLeNet have been widely used in image processing applications. However, these pre-trained networks are trained by non-medical images; non-medical images and medical images have significant differences. Therefore, the CNN models using transfer learning of these pre-trained networks may not be an optimized option for medical image processing. In this paper, our simple CNN model and the CNN models based on these pre-trained networks are applied for heart diseases image classification respectively. The simulation result demonstrates that our simple CNN model achieves higher image classification accuracy and less training time. Our paper provides a clue that a simple CNN model directly trained by medical images may be a better option for medical image processing, compared with the CNN models using transfer learning of the pre-trained networks trained by non-medical images.