Deep Learning Models with Applications to Brain Image Analysis

Abstract

Deep learning, rooted in artificial neural networks, has received increasing attention in the field of brain image analysis. In this chapter, the pre-processing steps for brain images and the fundamental concepts of deep neural networks are first introduced. After that, four typical types of deep neural networks used for brain image analysis are elaborated, including (i) convolutional neural networks (CNNs) and the variants (i.e., fully convolutional networks and U-net), (ii) recurrent neural networks (RNNs) and the variant (i.e., long short-term memory model), (iii) auto-encoder, and (iv) generative adversarial networks (GANs) and the variants (i.e., Pix2Pix GAN and CycleGAN), as well as their applications in brain image classification, segmentation, registration, and image synthesis/augmentation. In addition, several challenges and future research directions of deep learning in brain image analysis are also pointed out.