Jianfei Liu; Vivek Batheja; Pritam Mukherjee; Perry J. Pickhardt; Peter C. Grayson; Ronald M. Summers
Atherosclerosis, characterized by plaque buildup in arterial walls, is often studied in coronary arteries. However, it is less studied in the aorta and iliac arteries, despite their links with diseases like ischemic heart disease, stroke, and periph-eral vascular conditions. This work develops an automated method to segment and label plaques in the aorta, left, and right iliac arteries. Using an nnU-Net-based segmentation framework, plaques, arteries, and organs that potentially in-duce false plaque detections are first segmented. Plaques are then labeled according to the arteries in which they are located. Agatston scores are calculated on labeled plaques to assess atherosclerosis burden. The proposed method was validated using 50 internal contrast-enhanced and 60 external non-contrast CT scans. For both internal and external datasets, the plaque burdens computed from the automated segmentation were found to be strongly correlated with those computed from manual annotations (R2>0.8, except for the left iliac artery in the external dataset). These results sug-gest the potential of automated plaque labeling method for atherosclerotic plaque burden assessment in clinical usage.