Improved plaque detection and segmentation for atherosclerotic plaque burden assessment on non-contrast and contrast-enhanced abdominal CT scans: multi-institutional study

Atherosclerosis is a hardening process of the arteries caused by plaque buildup in and on the artery wall. It is the main cause of many cardiovascular diseases, such as heart attack, stroke, and peripheral artery disease. The risk of these diseases is correlated with the extent of atherosclerotic plaque burden. This work targets automatic calcified plaque detection and segmentation to assess the plaque burden. It is formulated as multi-class segmentation of plaque, aorta, body, and background on CT scans. The body class helps remove false positives caused by bright structures outside the aorta. The aorta class limits the search range. The state-of-the-art deep learning framework, nnU-Net, is used to train the multi-class segmentation model. Agatston scores calculated on the segmented plaque regions are used to assess the plaque burden. This work used three datasets with 678 non-contrast and contrast-enhanced abdominopelvic CT scans from two hospitals. The sensitivity of plaque detection was 81.6±10.6% and 81.5±12.2% on datasets from the two hospitals, respectively, while it was only 66.2±21.0% and 43.5±21.0% using a baseline approach (p<0.001). The Dice similarity coefficient of plaque segmentation using the proposed method was 81.1±15.3% and 82.0±8.0%, respectively. It was significantly higher than the 74.2±21.6% and 53.7±20.6% using the baseline method (p<0.001). Linear regression indicated that there was no significant difference between the Agatston scores from non-contrast and contrast-enhanced CT scans of the same patient. These promising results demonstrated that plaque burden can be assessed more accurately on both non-contrast and contrast-enhanced CT scans by the proposed method, potentially leading to improved atherosclerotic disease diagnosis.