Research On Feature Extraction, Selection And Classification Algorithms For Pulmonary CAD

Lung cancer is one of the most common malignant visceral tumors, and it is usually diagnosed in later period with lower survival rate, so early diagnosis is significant. At present, technology of CAD in medical imaging emerges continuously and develops rapidly, and has become a hot issue in this field. Lung CAD system can help doctors assess the medical images, improve diagnostic efficiency and reduce the burden on doctors. Studies show that CAD has played a positive role in early detection, early diagnosis, improving diagnosis accuracy and reducing misdiagnosis of lung cancer.Feature extraction, selection and classification algorithms for Lung CAD system are analyzed and researched in the thesis. In the research of feature extraction, characteristics of solitary pulmonary nodule are analyzed from the view of medicine,25 features including gray-scales, shapes and texture features are extracted, which consist of a complete feature space that can express medical signs fully.In the aspect of feature selection, feature space is optimized before identification in order to reduce the unnecessary computation and improve the performance of classifiers. Firstly, the simple genetic algorithm is analyzed in pulmonary nodule feature selection. Although the result greatly reduces redundancy and computing time, it damages the integrity of the feature space because it deletes many features directly. So GA based on PCA in feature selection is proposed in this thesis. Before GA selection, principal component analysis is used to get complete fusion characteristics. And experiment results show it reduces feature dimensions and achieves the optimization of the feature space.BP neural network and support vector machine algorithms are mainly researched for ROI classification. On the view of efficiency and accuracy, the performance of SVM classifier is better than BP neural network. However, the loss of missed diagnosis is larger than misdiagnosis in ROI identification, so the loss-equal SVM algorithm to prevent missed diagnosis is proposed in lung noudle detection. Experiment results show the loss-equal SVM classifier has good performance.