Segmentation of Individual Ribs from Low-dose Chest CT

Segmentation of individual ribs and other bone structures in chest CT images is important for anatomical analysis, as the segmented ribs may be used as a baseline reference for locating organs within a chest as well as for identification and measurement of any geometric abnormalities in the bone. In this paper we present a fully automated algorithm to segment the individual ribs from low-dose chest CT scans. The proposed algorithm consists of four main stages. First, all the high-intensity bone structure present in the scan is segmented. Second, the centerline of the spinal canal is identified using a distance transform of the bone segmentation. Then, the seed region for every rib is detected based on the identified centerline, and each rib is grown from the seed region and separated from the corresponding vertebra. This algorithm was evaluated using 115 low-dose chest CT scans from public databases with various slice thicknesses. The algorithm parameters were determined using 5 scans, and remaining 110 scans were used to evaluate the performance of the segmentation algorithm. The outcome of the algorithm was inspected by an author for the correctness of the segmentation. The results indicate that over 98% of the individual ribs were correctly segmented with the proposed algorithm.