During the Fall of 2024 and Spring of 2025, ReLU collaborated with Oslo University Hospital to explore the potential of AI in identifying abdominal aortic aneurysms (AAA) from CT scans. AAA is a dangerous condition where an enlarged aorta can rupture, often with fatal consequences. While the mortality rate is high for ruptures, early detection dramatically improves patient outcomes. However, AAAs are frequently missed in CT scans taken for unrelated reasons.

Our project aimed to build an automated pipeline capable of identifying AAAs from general abdominal CT scans, including those not specifically optimized for vascular imaging. We utilized state-of-the-art tools such as MONAI and 3D Slicer to develop a segmentation-based approach, where the aorta is first identified and its diameter measured to assess rupture risk.

We experimented with several approaches, including training a custom 2D U-Net model and applying MONAI’s Auto3Dseg low-code pipeline, as well as using public models like TotalSegmentator. In addition to model development, the team built tools for data preprocessing, visualization, and post-processing analysis of aortic measurements. Manual segmentation of AAA cases and infrastructure for handling sensitive data were also established.

Though access to labeled AAA data remained a major challenge, our efforts laid the foundation for future development. The next team will focus on completing the segmentation dataset, refining the models for aneurysm detection, and potentially deploying tools that support early AAA diagnosis in clinical workflows.