The amount of Epicardial adipose tissue correlates with the probability of diabetes, which can lead toorgan and nervous system complications. However, since, its diagnostic method is easily influenced by external factorsuch as diet, necessitating the use of a feature extraction network to extract epicardial adipose tissue from CT images.To realize that, segmentation of epicardium is crucial. Identifying the epicardium and its surrounding tissues on CTimages is challenging due to the low contrast and the thin boundary that the epicardium presents. Previous methodsemployed commonly used CNN network such as DenseUNet to detect epicardial lines by learning local features, butthese methods are unable to capture global information. In this paper, we propose MCNet consisting of MedVit-CNNencoder to enhance the local and global information. The network improved F1 score by more than 2.8% thancompared to conventional methods.
NETL
