Auto Whole Heart Segmentation from CT images Using an Improved Unet-GAN

Abstract

The development of deep learning is rapid, and convolutional neural network especially U-Net plays an important role in the medical image segmentation tasks, which is lack of data. Lots of models and methods are proposed to segment cardiac CT images. In this paper, we proposed a new network architecture. The network architecture is based on a traditional architecture called conditional generative adversarial network (cGAN), where R2U-Net acts as the generative network and FCN as the discriminative network. The performance of this model running on the dataset from MICCAI 2017 Multi-Modality Whole Heart Segmentation Challenge (MM-WHS 2017) is good.