Research On Automatic Extraction Of ROI And Computer-aided Detection For Medical Image

Computer-aided detection is a product which is the combination of computer science, cognitive science, engineering, mathematics and medicine, and it has brought many conveniences to the clinical diagnosis of disease. Computer-aided detection technology has improved the clinical diagnosis lever of the medical imaging, and provided more accurate information for the treatment. It has the great significance to the basic human body research and the development of medical devices, and it has become one of the hot researches in medical diagnosis, the application in medical imaging has become more and more extensive and showed a rapid change.How to find the main message in the many of redundant information is the deciding factor to process and analysis of the amounts of image information. Therefore, extract the region of interest quickly, accurately and segment the region is a necessary premise for the doctor who gives the accurate diagnosis. Generally speaking, image segment method can be divided into edge-based detection method which needs the processing of edge tracking, as well as region-based method, but its segment effective depends on the initial shape.The main study object in this thesis is the application of medical image computer-aided detection system using in the detection of lung nodules. The feature of pulmonary nodules in the CT image is not obvious, the shape and location is different, and it is easy adhesion with other organizations. Computer-aided detection system can increase the amount of detected lung nodules and reduce the number of missed nodules, which can assist the clinicians to distinguish the benign and malignant nodules. Generally, with help of a computer-aided detection system, the lung image can be processed more effectively to indict the suspicious lung nodules. In the process of detection, feature extraction and classification are the most important part, and it affects the entire detection system directly. Therefore, this thesis focuses on automatic detection of the ROI. This paper presents a new method which is FCM based on the model of visual attention image segmentation. It gets the local saliency map through the model of human visual attention, and then the FOA is used to initialize the clustering center and do the fuzzy C-means clustering, in order to achieve the extraction of the ROI, thus completing the automatic detection of pulmonary nodules. Simulation results show that the method used in computer-aided diagnosis can improve the effectiveness of medical diagnosis, and improving the detection of the rate of lesions, and it plays a positive effect on the reducing of misdiagnosis and missed diagnosis.