Segmentation of Coronary Arteries from X-ray Angiographic Images Using a Combination of K-Nearest Neighbor Clustering and Morphological Reconstruction Techniques

Abstract

Coronary angiography is an X-ray procedure used to examine the arteries of the heart. It provides information about the presence and severity of heart disease and helps doctors assess how well the heart is functioning. This study introduces a new technique for segmenting the coronary arteries in X-ray angiographic images using K-nearest neighbors clustering. The method involves separating thick and thin veins into distinct spaces and then merging them together. To eliminate noise and extract the vessel tree accurately, the algorithm employs morphological techniques like reconstruction, skeletonization, pruning, and dilation, as well as filters such as mean and convolution filters. The resulting segmented vessel tree contains several newly identified thin vessels that have low light intensity in the original image. The algorithm's efficacy is demonstrated by comparing the results with ground truth images. The evaluation criteria, including an accuracy of 0.9747, specificity of 0.9784, and sensitivity of 0.9049, indicate favorable performance compared to other methods. Additionally, the performance of this method is assessed using multiple lesions and instances of vessel blockages.