Research Of Pulmonary Nodule In CT Images Feature Extraction And Classification Method Based On SVM

The lung cancer is one of the most familiar malignant tumors, and it also has one of the lowest survival rates after the diagnosis. The major reason is the early detection of cancers is very difficult and cure rate is very low in advanced stage. In order to improve survival rates, tumor diagnoses and treatment in the early stage is the key and primary method. The imaging finding of lung cancer inchoate is the Solitary Pulmonary Nodules. So, the detection and recognize of solitary pulmonary nodules played a significant role in lung cancer diagnose.With the development of modern imaging technologies, the low dose CT scanning becomes one of the most effective methods to diagnose the lung cancer. Computerized Tomography is a new technology based on X-ray and computer technology. Computer Aided Detection system has offered strong support for the early diagnosis of lung cancer by processing the SPN CT images. It doesn’t only reduce the huge work of doctor, improve efficiency, and also makes the diagnose objective, improves the precision level. So, the computer-aided diagnostic, the extraction and classification of pulmonary nodules has a quite important theoretical value and realistic meanings. It is also the intention to research in this thesis.This paper analyzed the common image processing and arithmetic, and mostly researched on the feature extraction and classification.The main contents are summarized as five parts:The current situation both in China and abroad. Correlative knowledge of lung disease and Pulmonary Nodules. Pre-Processing Methods of the CT images. Feature extraction of image. Classification of the tumors. This thesis discussed the feature extraction and classification for Pulmonary Nodule in theory and practice. And it also processes the clinical CT images. The Pulmonary Nodule images are only classified into "benign" and "malignant", because the clinical images are limited and there is a great variety of the lung diseases. Besides, the research is in the beginning stage, the algorithm of feature abstraction and classification has to be improved. The emphasis and direction of study in the future is to build a complete images processing system to auto-extraction the ROI area.