29 March 2016 Segmentation of ribs in digital chest radiographs
Author Affiliations +
Ribs and clavicles in posterior-anterior (PA) digital chest radiographs often overlap with lung abnormalities such as nodules, and cause missing of these abnormalities, it is therefore necessary to remove or reduce the ribs in chest radiographs. The purpose of this study was to develop a fully automated algorithm to segment ribs within lung area in digital radiography (DR) for removal of the ribs. The rib segmentation algorithm consists of three steps. Firstly, a radiograph was pre-processed for contrast adjustment and noise removal; second, generalized Hough transform was employed to localize the lower boundary of the ribs. In the third step, a novel bilateral dynamic programming algorithm was used to accurately segment the upper and lower boundaries of ribs simultaneously. The width of the ribs and the smoothness of the rib boundaries were incorporated in the cost function of the bilateral dynamic programming for obtaining consistent results for the upper and lower boundaries. Our database consisted of 93 DR images, including, respectively, 23 and 70 images acquired with a DR system from Shanghai United-Imaging Healthcare Co. and from GE Healthcare Co. The rib localization algorithm achieved a sensitivity of 98.2% with 0.1 false positives per image. The accuracy of the detected ribs was further evaluated subjectively in 3 levels: "1", good; "2", acceptable; "3", poor. The percentages of good, acceptable, and poor segmentation results were 91.1%, 7.2%, and 1.7%, respectively. Our algorithm can obtain good segmentation results for ribs in chest radiography and would be useful for rib reduction in our future study.
© (2016) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Lin Cong, Lin Cong, Wei Guo, Wei Guo, Qiang Li, Qiang Li, } "Segmentation of ribs in digital chest radiographs", Proc. SPIE 9788, Medical Imaging 2016: Biomedical Applications in Molecular, Structural, and Functional Imaging, 97881T (29 March 2016); doi: 10.1117/12.2216742; https://doi.org/10.1117/12.2216742

Back to Top