Automatic Classification of Calcification in the Coronary Vessel Tree

This paper presents a pipeline for complete automatic detection and labelling of the coronary artery calcium lesions. The pipeline not only distinguishes between coronary artery calcifications and non-calcifications but also labels the detected calcium lesion to one of the main coronary vessel tree. The pipeline presented uses a combination of atlas-based and machine learning approaches. The set of features supplied to the classifier are either directly obtained from the subject scans or are derived after using information from atlas scans. The classifier was trained using 76 CT datasets. The pipeline was tuned using the training set provided by the challenge evaluation framework. On the 32 test datasets we achieved a sensitivity of 67% and a PPV of 85%, comparing the volume of the detected calcifications it was 89% and 94% respectively. The ICC was 0.99 for LAD+LM, 0.98 for LCX and 0.92 for RCA, on the whole heart the ICC for Agatston and volume score were 0.99 and 0.98 respectively. An accuracy of 88% and a linear weighted kappa (κ) of 0.90 was achieved for the risk categorization accuracy.