The Research Of Pre-Computer-Aided Detection Of Lung Cancer Based On CT Images

Since 70s of the last century, morbidity and mortality of lung cancer have been rising sharply in the world, while the number of patients with lung cancer in China is the world's first. Clinical studies have shown that if lung cancer can be detected early and treated, the survival rate of five-years will increase from 14% to 49%. What's more, 90% of lung cancers at an early stage are likely to be found, but up to 30% of the lung cancer focus may be leaved out when the doctor diagnoses lung CT images alone. Therefore, in order to improve the early detection of lung cancer, studies of pre-computer-aided detection of the focus of lung cancer based on CT images have a very important practical significance on human life and health. The research has important theoretical significances, as it is a cross interdisciplinary research, which comes down to medical imaging, medical image processing, artificial intelligence and pattern recognition and many other disciplines.In this paper, on the basis of relevant research at home and abroad,combining the technology of attribute reduction based on the rough set theory with fast classification for support vector machines, the rapid detection method is proposed based on RS_FCSVM, which is used in the pre-computer-aided detection of the focus of lung cancer. The method improves the effectiveness and feasibility of lung cancer detection significantly.In this paper the main research work includes:(1) According to the characteristics of lung CT images, this paper adopts the method of the global adaptive threshold segmentation, contour tracing and connected component labeling to obtain a complete lung parenchyma; Using K-means clustering method to extract the region of interest of the focus of lung cancer.(2) According to the medical signs of the focus of lung cancer, the original feature set of detection of the focus of lung cancer is extracted out. An effective feature selection method of detection of the focus of lung cancer based on the rough set theory is proposed. The experimental results show that the method can effectively remove redundant features, and would construct a streamlined and effective feature set of detection of the focus of lung cancer.(3) On the basis of the effective feature selection of detection of the suspicious focus of lung cancer based on the rough set theory, a rapid detection method is proposed based on RS_FCSVM, which is used in the pre-computer-aided detection of the focus of lung cancer. The experimental results show that the method has achieved the desired results, reducing the dimension of input samples and the complexity of detection, improving accuracy, timeliness and efficiency of the pre-computer-aided detection of the focus of lung cancer.(4) According to the above method, a set of user-friendly, operation-easy application software of the pre-computer-aided detection of the focus of lung cancer is developed, using Visual C++ 6.0 in Windows XP.