Automated coronary calcium scoring using predictive active contour segmentation

Agatston scoring has become the primary clinical method of calcified lesion quantification for the diagnosis of coronary artery disease. Such a method compares a patients coronary plaque burden against a series of preset disease severity ratings. This burden allows a cardiologist to make a risk assessment of future cardiac events and prescribe suitable treatment. The main factor affecting plaque quantity lies with a major aspect of user interaction, the manual selection of potential calcified lesion ROI´s. Being a manual process, the issue of subjective ROI selection and reproduction by different experts can lead to varying results for a single patient. Thus an automation of the scoring process has been proposed. In order to test this, an automated method of plaque lesion detection and scoring has been presented, validated against corresponding clinical calcium scores. This approach has shown as good results as obtained clinically with the exception of certain outliers caused by either over or under inclusion of lesions.