Abstract

Cardiovascular disease (CVD) is the leading cause ofmortality yet largely preventable, but the key to preventionis to identify at-risk individuals before adverse events. Forpredicting individual CVD risk, carotid intima-media thick-ness (CIMT), a noninvasive ultrasound method, has provento be valuable, offering several advantages over CT coro-nary artery calcium score. However, each CIMT exami-nation includes several ultrasound videos, and interpretingeach of these CIMT videos involves three operations: (1)select three end-diastolic ultrasound frames (EUF) in thevideo, (2) localize a region of interest (ROI) in each selectedframe, and (3) trace the lumen-intima interface and themedia-adventitia interface in each ROI to measure CIMT.These operations are tedious, laborious, and time consuming, a serious limitation that hinders the widespread utilization of CIMT in clinical practice. To overcome this limitation, this paper presents a new system to automate CIMT video interpretation. Our extensive experiments demonstrate that the suggested system performs reliably. The reliable performance is attributable to our unified framework based on convolutional neural networks (CNNs) coupled with our informative image representation and effective post-processing of the CNN outputs, which are uniquely designed for each of the above three operations.