A Detection Algorithm Based On Morphology And Gray Entropy For Pulmonary Nodules

In recent years, lung cancer is one of the biggest malignant tumors which have most damaging to human life. The main mean for the detection of early lung cancer is using imaging methods in medical diagnosis, including X-ray diagnosis and CT diagnosis, and for higher detection rate of CT diagnosis, this kind of method get more extensive application. Lung cancer most commonly manifests as pulmonary nodules which are visible on CT scans. They have various shapes, different sizes, uncertain locations and the gray values of pulmonary nodules are similar to those of blood vessels and bronchus in the lung. It is difficult to distinguish the pulmonary nodules and soft tissues in the lung even by experienced doctors. In addition, radiologists'diagnosis is also a kind of subjective judgment process, so in this diagnosis process the doctor will be limited by their own experience, knowledge level and external conditions and even with the same doctor will also produce reading differences to a same slice in different times. Accordingly, computer aided detection and diagnosis (CAD) is developed, and it can analyze CT images of lung automatically and present pulmonary nodules to help doctor more effectively analyze data security and overcome some objective factors.After researching on the differences between pulmonary nodules and bronchus, pulmonary vessels in the CT scans, this paper presents a detection algorithm based on morphology and gray entropy for pulmonary nodules. This algorithm is mainly based on the multi-scale filtering of morphology and the selecting of gray entropy. The main work are as follows:(1) According to the different geometry shapes of pulmonary nodules in CT image, three circle-like structure elements with different dimensions are built, and the multi-scale morphologic filtering is adopted to get the initial candidates of the regions of interest (ROI). After this processing, the circular regions which are bigger than the structure elements are enhanced while the linear regions including bronchus and blood-vessels are suppressed; then, in order to further reduce the number of ROI and increasing the speed of the operation, according to the gray variation differences between pulmonary nodules and bronchus and blood-vessels, gray entropy is adopted to distinguish the pulmonary nodules from others.(2) In the feature selection section, considering the pulmonary nodules on the gray, shape, texture feature and choose the Ave, Var, S, C, FD, CD as the feature to compute the regional features of candidate nodules; then, the candidate nodules are classified by the nearest neighbors classifier and SVM classifier. The last, the corresponding regions of candidate nodules sentenced pulmonary nodules are marked in the surface of the software.Experiment results indicate that the detection algorithm based on morphology and gray entropy for pulmonary nodules can extract suspected pulmonary nodule regions in the CT images effectively, which is a basis for subsequent pulmonary nodule identification and diagnose.