Abstract 4371086: Population-scale AI Measurement of the Aorta Identifies Novel Traits Associated with Aortic Diameter

Abstract

Introduction: Aortic diameter is influenced by age, sex, body surface area (BSA), and race. Studies of aortic diameter have been limited due to small sample size or focus on specific diseases such as aortic aneurysms. Hypothesis: Population-level measurement of aortic diameter can reveal determinants of aortic diameter and aneurysm. Methods: Using a novel AI approach, we automatically measured the maximum and mid-point diameters in the ascending, descending, and abdominal aorta from CT scans among 187,465 patients aged 20-85 years (56% female). Z-scores for aortic diameter were created using an XGBoost model, adjusting for age, sex, BSA, and race. Linear regression was applied in a comprehensive association test for 1,857 clinical traits gathered from electronic medical records against the Z-scores. Bootstrap analysis using a linear mixed model was performed to allow robust estimation of effect size on aortic diameter for select traits. Results: The average abdominal aortic diameter in our population was 18±5mm. The average descending aortic diameter was 25±4mm. Our analyses confirmed positive association of aortic diameter with male sex, age, and BSA as well as known clinical risk factors including, but not limited to, hypertension, history of smoking, and atherosclerosis (p-values < 0.05). Remarkably, we found a strong negative association between developmental delay and diameter Z-score, with standardized coefficients of -0.465 (SE 0.053, p=1.95×10 -18 ) for the abdominal aorta and –0.35 (SE 0.045, p=4.81×10 -14 ) for the descending aorta. We used 1:5 unique case-control matching on age, sex, smoking status, and BSA for 1,119 patients with a diagnosis for developmental delay. When compared to controls, the average difference in abdominal aortic diameter in this group was -1.16±0.046mm (95% CI: -1.07 to -1.25 mm). For comparison, the average increase in abdominal aortic diameter amongst those with a history of smoking was 1.17±0.093mm (95% CI: 1.00 to 1.36). Conclusions: Population-scale analysis of aortic diameter confirmed the known demographic and diagnostic risk factors for aortic enlargement. Furthermore, we identified a significant negative association of developmental delays with abdominal and descending aortic diameter, suggesting a novel biological link.