ABSTRACT
The enormous advancement of Deep Learning (DL) algorithms in various fields has been adopted by many perspectives of medical image analysis. In the context of deep learning, convolutional neural networks (CNNs) are widely used models for medical imaging tasks and lead to a huge progress in computer aided diagnosis. CNNs are being profoundly absorbed in the research of medical imaging because of their pertinent features in preserving local image relations and dimensionality reduction. CNNs demonstrate excellent performance in lesion detection and other vision tasks. This chapter presents the modules of convolutional neural networks and their prominent architectures employed for medical image analysis. The commonly used variants of CNNs in medical image analysis including ResNet, GoogleNet,
and fully convolutional neural networks are analyzed. The principal research areas and applications of medical image registration, segmentation, detection and classification are discussed. The overview of the above mentioned tasks in a few of the usual diagnosis areas of the human body such as brain, lungs and eye is presented in this chapter. The crucial challenge in medical image analysis using CNNs is the availability of training data sets. This problem is addressed in this chapter, by providing the existing medical image datasets for different application areas like chest, diabetic retinopathy, lungs and heart. Finally, the chapter is concluded by discussing the challenges associated with CNN architectures and promising future research directions.
