Segmentering av hjertekammer i CT bilder, uten bruk av kontrastvæske

Offentlig prosjektsammendrag

This pre-project addresses critical challenges in cardiac computed tomography (CT) imaging, focusing on the harmful effects of contrast agents used in CT angiography. The project will develop accurate DL models for segmenting cardiac structures in NCCT. The project's significance extends to both human healthcare and animal husbandry. In the context of pig farming, sudden death losses due to heart failure are a concern, necessitating robust cardiovascular traits in breeding. A unique dual CT dataset collected and developed by Norsvin, comprising CCTA and NCCT images, forms the basis for this research, emphasizing the need for accurate 3D segmentation. The project will employ a novel approaches, integrating Swin Transformer (ST) and Convolutional Neural Network (CNN) backbones. Such architectures enables global information capture and local feature extraction, addressing the challenge posed by similar intensities in NCCT images.