Deep learning-driven false-lumen volumes predict adverse remodeling better than diameter in patients with residual aortic dissection on CT

Question In residual aortic dissection (RAD) after type-A dissection, early identification of high-risk patients on initial CT angiography is crucial for endovascular treatment decisions. Findings False-lumen local volumes (3 cm around aortic dissection maximal diameters), obtained with an automatic deep-learning method, predict adverse remodeling better than diameter or global false-lumen volumes. Clinical relevance A deep-learning segmentation method of aortic dissection components on CTA, enabling automatic measurements of diameters and volumes is feasible. It provides local false-lumen volumes, a better predictive marker of adverse aortic remodeling than the currently used diameters and global volumes.