The Study Of Detection Arithmetic For Pulmonary Opacity Based On CT Image

Lung cancer is a common malignant tumor, which is one of the five biggest threats to human life and health. In the early stage of lung cancer, patients have no obvious symptoms, so it is often neglected. When patients cough, haemoptysis, chest pain and other clinical symptoms, they are usually collected chest CT images, but the tumor found is too late now. If lung cancer can be found earlier, patients can be treated earlier, thus the cure will increase. Early lung cancer usually exists as pulmonary nodules, so checking the pulmonary nodules earlier, lung cancer can be diagnosed and treated earlier.The Computer Aided Diagnosis of pulmonary nodules problem is studied in this thesis. The main aspects of research include:(1)Segment the lung areas. Due to detect pulmonary nodules is conducted on the basis of lung parenchyma segmentation, so the CT value of lung can not destroied. In view of this goal,this article mixed has used the region growing, the binaryzation and the threshold way to segment the lung areas, and getting the satisfactory division result.(2)Extract the Region Of Interest. This paper proposed a mix algorithm, with which the region of interest can be extracted accurately. When extracted the region of interest, the extracted massive non-lung tubercle regions often affects the following classified result. In view of this question.This paper proposed a method about examines the lung tubercle by using Top-hat and the Gabor filter's mix algorithm, there are the good effect in reducing the false positive aspect by guarantee feels the interest region be completely extracted.(3)Classifying the abnormal areas. Owing to a great deal of false positives lie in ROI, the SVM classifier is designed to distinguish nodules from normal areas with good detected result after the effective features extraction. This article has chosen the specificity obvious 8 characteristics through compare the specificity of the choosed characteristic, including 4 shape characteristics,2 gradation characteristics, with 2 position characteristics.And aims at the extraction of the interested region is the small sample characteristic, using the SVM classifier. According to Classifying the ROI of the choosed characteristics has obtained the good classified effect.The results is satisfying which have been tested by processing the clinical image sets by emulator. And the sensitivity is 0.9436, the undetected rate is 0.0563. Therefore this article proposed the algorithm can good satisfy doctor's demand standard.