Automating Carotid Intima-Media Thickness Video Interpretation with Convolutional Neural Networks

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