Multi-stage learning for segmentation of aortic dissections using a prior aortic anatomy simplification

Aortic dissection (AD) is a life-threatening cardiovascular disease with high mortality rate. The accurate and generalized 3-D reconstruction of AD from CT-angiography can effectively assist clinical procedures surgery plans, however, clinically unavaliable due to the lacking efficient tools. In this study, we presented novel multi-stage segmentation framework for type B extract true lumen (TL), false (FL) all branches (BR) as different classes. Two cascaded neural networks were used segment aortic trunk separate dual lumen, respectively. An straightening method was designed based on prior vascular anatomy AD, simplifying curved shape before second network. straightening-based achieved mean Dice scores 0.96, 0.95 0.89 TL, FL, BR multi-center dataset involving 120 patients, outperforming end-to-end multi-class methods without dual-lumen segmentation, even using network architectures. Both global volumetric features aorta local characteristics primary tear could be better identified quantified straightening. Comparing previous deep learning dealing segmentations, proposed advantages in accuracy.