Thoracic Aortic Three-Dimensional Geometry

Area of Science:

• Cardiovascular Imaging
• Biomedical Engineering
• Radiology

Background:

• Aortic structural degeneration, linked to aging, increases cardiovascular risks by affecting left ventricular afterload and arterial pulsatility.
• Comprehensive characterization of three-dimensional (3D) aortic geometry in large populations is lacking.
• Understanding aortic structure is crucial for cardiovascular health assessment.

Purpose of the Study:

• To develop and deploy an automated deep learning method for comprehensive 3D thoracic aorta segmentation.
• To extract multiple aortic geometric phenotypes (AGPs) across diverse aortic subsegments.
• To enable large-scale studies on the clinical implications of aortic structural changes.

Main Methods:

• A deep learning architecture was utilized for complete thoracic aorta segmentation.
• Morphological image operations were employed to derive AGPs, including diameter, length, curvature, and tortuosity.
• The method was applied to imaging data from 54,241 UK Biobank participants and 8,456 Penn Medicine Biobank participants.

Main Results:

• A fully automated approach for quantifying 3D aortic structural parameters was established.
• Aortic geometric phenotypes were expanded across two large, representative biobanks.
• The study successfully segmented the complete thoracic aorta in a large cohort.

Conclusions:

• The developed automated method facilitates the quantification of 3D aortic geometry.
• This approach enhances the available phenotypic data for large-scale cardiovascular research.
• It will aid in elucidating the biology and clinical consequences of aortic degeneration in aging and disease.