CT angiography scans (n=40) of patients with TBAD were analyzed. The true and false lumen were segmented by a radiologist with 8 years of experience in cardiovascular imaging using commercially available software (Intuition, TeraRecon, Foster City, CA). The SAM model was integrated into an analysis script developed in Python (v3.10.10, 2023), enabling per-slice true and false lumen segmentation with minimal user-input (typically up to 10 mouse-clicks). Segmentation accuracy was assessed through Dice scores. Volume comparisons were made using linear regression and Bland-Altman analyses. The volumes of the true and false lumen derived from the manual and SAM segmentations were tested for statistically significant differences using the two-sample t-test.

Fig 1: Overview of the segmentation procedure using Meta AI's Segment Anything Model...