Towards Full Automation of Geometry Extraction for Biomechanical Analysis of Abdominal Aortic Aneurysm; Neural Network-Based versus Classical Methodologies

In this study we investigated the impact of image segmentation methods on the results of stress computation in the wall of abdominal aortic aneurysms (AAAs). We compared wall stress distributions and magnitudes calculated from geometry models obtained from classical semi-automated segmentation versus automated neural network-based segmentation. Ten different AAA contrast-enhanced computed tomography (CT) images were semi-automatically segmented by an analyst, taking, depending on the quality of an image, between 15 and 40 minutes of human effort per patient. The same images were automatically segmented using PRAEVAorta 2, commercial software by NUREA (https://www.nurea-soft.com/), developed based on AI algorithms, requiring only 1-2 minutes of computer time per patient. Aneurysm wall stress calculations performed using the BioPARR software (https://bioparr.mech.uwa.edu.au/) revealed that, compared to the classical semi-automated segmentation, the automatic neural network-based segmentation leads to equivalent stress distributions, and slightly higher peak and 99th percentile maximum principal stress values. This difference is due to consistently larger lumen surface areas in automatically segmented models as compared to classical semi-automated segmentations, resulting in greater total pressure load on the wall. Our findings are a steppingstone toward a fully automated pipeline for biomechanical analysis of AAAs, starting with CT scans and concluding with wall stress assessment, while at the same time highlighting the critical importance of the repeatable and accurate segmentation of the lumen, the difficult problem often underestimated by the literature.