Integration of AI and Machine Learning in Predicting Outcomes of Endovascular Aneurysm Repair

Abstract

One of the most common side effects following endovascular aortic repair (EVAR) is endo-leaks, which can result in subsequent rupture and higher rates of reintervention. Cross-sectional imagination with manual axile dimensions has historically been used for the necessary serial lifetime monitoring. Imaging analysis based on artificial intelligence (AI) has been created and might offer a quicker and more accurate evaluation. The goal of this study is to determine an AI-based program's ability to detect endo-leaks, link them with EVAR-related bad outcomes, and evaluate post-EVAR morphological modifications over time. Patients who had EVAR at a tertiary hospital between January 2017 and March 2020 and had at least two follow-up computed tomography angiography (CTA) assessments were evaluated using PRAEV Aorta 2 (Nurea). The program was tested against the ground truth supplied by human experts utilizing Sensitivity (Se), Specificity (Sp), Negative Predictive Value (NPV), and Positive Predictive Value (PPV). Adverse events associated with endovascular aortic repair were characterized as limb occlusion, endo-leak, rupture, aneurysm-related mortality, and EVAR-related re-interventions. A thorough anatomic description of aorta remodeling following EVAR was made possible by the AI-based program PRAEV Aorta, which also demonstrated interest in automatically detecting endo-leaks during follow-up. When compared to maximal diameter, the correlation between aortic aneurysmal volume and EVAR-related adverse events and endo-leaks was stronger.