Abstract Body (Do not enter title and authors here): Background
Organ size and body size relations are often non-linear (allometric). However, allometric relations for thoracic aorta are poorly understood and current practice is based on linear (ratiometric) indexation of aortic dimensions, which can provide misleading information in individual patients. Assessing allometric relationships can have important implications for the detection of aneurysms, particularly at the extremes of body size.
Methods
A U-Net convolutional neural network (CNN) segmentation architecture was trained, validated, and used to segment the ascending aorta (AA) and derive its maximum cross-sectional dimensions using MRI scans obtained from 49,193 UK Biobank participants. Sex-adjusted allometric (log-log) models were used to derive allometric exponents describing the relationship of AA size (diameter and cross-sectional area) to body size (height and BSA) in a healthy reference subgroup (n=3,972). Prognostic associations with all-cause mortality and aneurysm complications were then assessed in the entire population over a median follow-up of 66 months.
Results
Relationships between AA diameter or area and BSA were highly non-linear (Table 1) and exhibited significant differences between males and females (P for interaction=0.013), precluding the use of a single exponent for both sexes. In contrast, a quasi-linear relationship was found for aortic area and body height (exponent=0.98), making simple ratiometric indexation allometrically adequate. Moreover, no differences in this allometric relation were found between men and women (P for interaction=0.32), indicating its suitability regardless of sex. In contrast, ratiometric diameter/height, diameter/BSA, or area/BSA indexation provided markedly and systematically biased prevalence estimates for aortic dilation, misclassifying many individuals at the extremes of body size (Figure). Finally, the use of the linear area/height ratio was consistently predictive of death and aneurysm complications in the overall population and provided the highest C-index for aneurysm complications regardless of body size (Table 2).
Conclusions
In contrast to all other indexation methods, the linear AA area/height ratio is a practical, allometrically correct, and most informative aortic size index, and can be utilized equally in men and women. This should guide strategies for aneurysm detection and decision making in clinical practice and future studies.