Automatic vessel wall segmentation of IVOCT images using region detection EREL algorithm

Abstract

Intravascular optical coherence tomography IVOCT is a catheter-based imaging modality that uses near-infrared light, to produce high-resolution cross-sectional images of the vessel wall. Segmentation of the vessel wall is important to indicate stenosis and analyze atherosclerotic plaques. In this study we use the recently proposed region detector, named Extremal Region of Extremum Level (EREL), to detect the lumen and media contours in IVOCT frames, and then we used a region selection method to detect the most precise lumen and media contours from the extracted ERELs. We evaluated the proposed method on the dataset containing 142 IVOCT images. We get, the average Hausdorff Distances (HD) and Dice metric (DSC) between the extracted ERELs and the lumen and media contours, 0.045 mm, 0.141 mm and 0.986, 0.96, respectively. The results of our study showed that the IVOCT image segmentation using the proposed method is more robust and more precise than state-of-the-art.