Whole Heart Auto Segmentation of Cardiac CT Images Using U-Net Based GAN
Zeyu Lou, Weiliang Huo, Kening Le, Xiaolin Tian · 2020
The whole heart segmentation of medical CT images is of great significance for assisting doctors in the diagnosis of cardiovascular diseases and guiding doctors' surgery. Due to the complexity and particularity of medical images, automatic whole heart segmentation still remains challenges. Lack of annotated medical images is also a big problem. The U-Net successfully solved the problem of inadequate annotated medical images. In this study, we proposed a U-Net based GAN which uses U-Net as the generative network and FCN as the discriminative network. The experiments were performed on the dataset of the MICCAI 2017 Multi-Modality Whole Heart Segmentation Challenge (MM-WHS 2017). The proposed method achieved high segmentation accuracy with an average 86.32% and highest 93.64% Dice similarity coefficient (DSC) for the whole heart segmentation.