Inner Diameter Measurement Oriented Aortic Segmentation: An Edge Enhancement and Contextual Fusion Deep Learning Method

作者：Di Zhang, Wenjing Zhang, Tao Luo, Ming Yang, Aijun Liu · 年份：2023 · DOI：10.1145/3586139.3586149 · 被引用次数：1 · 研究领域：Advanced X-ray and CT Imaging、Medical Image Segmentation Techniques、Advanced Neural Network Applications

Coarctation of aorta (CoA) is a critical congenital malformation which can lead to serious complications such as hypertension, heart failure and even shock in severe case. The effective diagnosis and surgical management of CoA require accurate aortic inner diameter measurement mainly based on cardiovascular computed tomography (CT). Due to the traditional manual aortic inner diameter measurement is labor-intensive and susceptible to observer’s expertise, the deep learning (DL) enabled aortic segmentation based aortic inner diameter measurement methods have been investigated. However, most existing DL based aortic segmentation methods ignore the edge information and spatial consistency, which lead to poor segmentation performances. To solve this problem, we propose an edge enhancement and contextual fusion network, called ECN, which can enhance edge information and utilize the contextual relationships of CT slices so as to improve aortic segmentation. Simulation results show that the proposed algorithm outperforms the compared DL algorithms in dice score (0.9370) and 95% Hausdorff distance (1.3383mm) in our private patient-specific CoA dataset. Moreover, the proposed ECN based aortic inner diameter measurement achieves low bias and high correlation with the results measured by doctors.