Automatic coronary artery detection in cardiogram image

Abstract

The Myocardium Infarction or generally known as MI, is the most dangerous and serious disease that effects the health of the human heart. MI usually attacks and damages the heart muscle by blocking the path of coronary artery. Echocardiography is a widely used imaging technique to examine myocardial function in patients with known or suspected heart disease. In clinical practice, the analysis mainly relies on visual inspection or manual measurements by experienced cardiologists. For researchers, the analysis of myocardium can be done with tissue samples taken from very small areas of interest manually. However, this manual process is tedious and time-consuming, and is vulnerable to error due to human weaknesses. Therefore, this project is aimed to automatically detect the wall of the coronary artery that separates the left and right chambers of the myocardium, so that it becomes much easier to analyze this particular area. In order to achieve this aim, a number of per-processing methods have been applied on the real medical image of the heart from the echocardiography which is performed in an offline manner. These pre-processing operations involved the use of binarizing, down sampling, median filtering, morphological erosion, morphological dilation, logical AND operation and morphological connected component labeling. The classification of the coronary artery area was done based on two features extraction. One is the height to width ratio and another is the label to image area ratio. The performance of the proposed method achieved 90% accuracy based on 10 test samples