An automated convolutional neural network algorithm with human-in-the-loop validation for tracking thoracic aortic diameters at CT angiography demonstrated stronger agreement with reference standard vascular deformation mapping than manual clinical measurements and may support more efficient longitudinal surveillance.