AAA-100: A Curated Dataset of 3D Watertight Abdominal Aortic Aneurysm Models

<p>An abdominal aortic aneurysm (AAA) is a local dilatation of the abdominal aorta exceeding 30 mm that might rupture, with fatal outcomes in 70-80% of cases. Personalized 3D models of AAAs, including surrounding vasculature such as iliac and renal arteries play an important role in tailored clinical decision-making for AAA patients. Models could be used for, e.g., AAA growth modeling, stentgraft sizing and positioning for endovascular aorta repair (EVAR) procedures, or 3D printing for surgical&nbsp;practice. Extracting high-quality 3D arterial models from imaging modalities such as computed tomography angiography (CTA) is a time-consuming and challenging problem. For downstream applications such as computational fluid dynamics (CFD) or shape analysis, models should have sub-voxel accuracy, be watertight, and adhere to topological constraints. We present the AAA-100 dataset, containing 100 detailed 3D AAA models with consistent anatomical boundaries acquired semi-automatically from pre-operative CTA scans. These models span a wide range of possible AAA pathology. Moreover, all models are carefully curated to be anatomically and topologically correct.&nbsp;</p> <p>A detailed description of the data set and file structure is provided in description.pdf.</p> <p>We kindly ask you to cite the following works when using the AAA-100 dataset in your research</p> <blockquote> <p>Alblas, D., Suk, J., Brune, C., Yeung, K. K., &amp; Wolterink, J. M. (2025). SIRE: Scale-invariant, rotation-equivariant estimation of artery orientations using graph neural networks.&nbsp;<em>Medical Image Analysis</em>, 103467.</p> <p>Rygiel, P., Alblas, D., Brune, C., Yeung, K. K., &amp; Wolterink, J. M. (2024). Global Control for Local SO (3)-Equivariant Scale-Invariant Vessel Segmentation.&nbsp;<em>arXiv preprint arXiv:2403.15314</em>.</p> </blockquote>