Automated Deep Learning Pipeline for Accurate Segmentation of Aortic Lumen and Branches in Abdominal Aortic Aneurysm: A Two-Step Approach

Abdominal Aortic Aneurysm (AAA) is a serious medical condition characterized by the abnormal enlargement of the abdominal aorta. If left untreated, AAA can have life-threatening consequences. Accurate segmentation of the aorta in Computed Tomography Angiography (CTA) images plays a vital role in treatment planning for AAA. However, manual and semi-automatic segmentation methods suffer from limitations in terms of time and accuracy. This study presents a deep learning pipeline that aims to fully automate the precise and efficient segmentation of the aorta and its branches within CTA images. A two-step approach is proposed for the segmentation of the aorta and its branches in contrast-enhanced CTA scans. The first step involves utilizing a UNet++ model to perform aortic segmentation across the entire set of CTA axial slices. In the second step, a post processing algorithm is employed to track the continuity of the segmented aorta while effectively discarding false positive (FP) objects. The assessment of the fully automated method revealed remarkable outcomes, with a mean Dice coefficient of 0.941 on a test set consisting of 10 CTA scans. The automated segmentation results are utilized to create a comprehensive 3D model. The study results indicate that the utilization of the proposed deep learning-based pipeline is highly effective in achieving accurate segmentation of the aortic lumen and its branches. The practical implications of this approach extend to pre-operative planning. This highlights the valuable contribution of the proposed method in improving the management and treatment of patients diagnosed with aortic aneurysm.