Automated pericardium segmentation and epicardial adipose tissue quantification from computed tomography images

Ying Wang, Ankang Wang, Lu Wang, Wenjun Tan, Lisheng Xu, Jinsong Wang, Songang Li, Jinshuai Liu, Yu Sun, Benqiang Yang, Steve E. Greenwald

Abstract: Highlights•The BMT-UNet combines BE, multi-scale feature extraction, and ConvT modules to improve pericardial boundary segmentation.•Multi-scale extraction captures global and local info, while morphology operations address segmentation gaps.•A weighted multi-slice strategy uses adjacent slices to preserve the 3D pericardium, aiding EATV analysis and detection.