Capnext: Unifying Capsule And Resnext For Medical Image Segmentation

Capsule Network is a new approach to image analysis that emphasizes objeCT-part relationships. However, their applications to segmentation tasks are limited due to the difficultyto train to them. In this study, we proposed a novel Capsule Network, called CapNeXt, that unifies Capsule and ResNeXt architectures for medical image segmentation. Our approach advances the existing Capsule Network-based segmentationby making it possible to integrate techniques in Convolutional Neural Networks (CNN) to Capsules architecture. By doing so, our architecture is much easier to train compared with other contemporary approaches using Capsule in segmenta-tion tasks. We validate our approach in two public data sets. The experimental results show that CapNeXt outperforms the CNNs and other Capsule architectures in the segmentation task by at least 1% of the Dice score. The code will be released on GitHub after acceptance.