Emergence of Capsule Network for Automatic Medical Disease Classification

Modern methods for classifying Medical diseases make use of deep convolutional neural networks (CNNs). To achieve strong generalization, these networks need a lot of training examples since they are ineffective at encoding the part-whole links. In place of CNNs, recent studies suggest an effective variation of capsule networks (CapsNets). Automatic medical disease classification using images of different modalities has received significant attention recently. We conduct the systematic study of CapsNets for automatic medical disease classification using medical images. We present the architecture and design of CapsNets with its different phases. The review of recent CapsNets-based studies presented a comparative analysis. The outcome of this paper is the current progress of CapsNets for medical disease classification and its challenges.