Lung Parenchyma And Ribs Segmentation For Computed Radiography

Along with the development of technology and the improvement of standard of living , the incidence of lung cancer increases year after year in China. Computer Aided Detection (CAD) system is of importance and very useful for detection and diagnose of lung cancer in physical examination. As a important part of lung CAD, the accuracy of the lungs and ribs segmentation is very crucial.CR (Computed Radiography) is a common screening procedure with little radiation. Now chest radiography occupies almost forty percentages in imaging methods. However, because of the lower contrast between the tissue, lungs and ribs segmentation in chest radiograph is very difficult. In this paper an effective method was presented. First, a new automatic algorithm for the segmentation of lung fields in chest radiographs is proposed, which is used as the first layer of a hierarchical segmentation based on OTSU and adaptive thresholds. In the hierarchical algorithm, the properties of the contrast of air and muscle is made use to initially segment the air background and the parenchyma. Then we use knowledge of the position of air background to separate the air background and parenchyma completely. The experimental results verify the effectiveness of this algorithm. Furthermore, a novel Guassian unsharp masking thretholding segmentation algorithm of the ribs in chest radiographs is proposed. First, we use Guassian unsharp masking threthold algorithm to segment the candidate regions of ribs. Since there are many noises and tracheas among the candidate ribs, we use parabolas fitting to extract the rib borders. The experiments demonstrate the resultant rib borders are coincided with the ground-truth..The methods in this paper are valid in about 100 chest radiographs.