Calcification Detection Using Deep Structured Learning in Intravascular Ultrasound Image for Coronary Artery Disease

Coronary artery disease (CAD) is also known as atherosclerosis, a non-communicable disease (NCD) in cardiovascular disease (CVD). The plaques and the calcification embedded in the coronary artery inner wall make the blood vessel area narrow. The standard practice by the radiologists and medical clinical are by visual inspection to detect calcification on the intravascular ultrasound (IVUS) image. In this study, we focus on detecting the calcification present and calcification absent in the coronary artery disease for catheter frequency 20MHz IVUS image using Directed Acyclic Graph (DAG) network. The four types images that are Cartesian coordinate image, polar reconstructed coordinate image, Cartesian coordinate warp image with and polar constructed warp image of were used as an input image in this study. The convolutional neural network (CNN) using Directed Acyclic Graph network (ResNet101) together with the three types classifier (Decision Tree, K-nearest Neighbour and Naïve Bayesian) were investigated for the detection of the calcification present and calcification absent. In this work, we used dataset B from MICCAI Challenge 2011 that consists of 2175 coronary artery disease IVUS image where 530 of the images with calcification and 1645 are without calcification to demonstrate our framework. The performance measures such as the accuracy, sensitivity, specificity, positive prediction value and negative prediction value were evaluated and compared with the ground truth provided. The ResNet101 architecture, the four types of images and the three types of the classifier gave good accuracy.