Abstract: In order to improve the accuracy of benign and malignant diagnosis of the solitary pulmonary nodules in the computer aided diagnosis system,this paper proposed a novel classification algorithm based on genetic algorithm and back propagation neural networks.Considering the local optimum problem of the BP neural networks and the medical diagnosis features of solitary pulmonary nodules,the proposed algorithm uses genetic algorithm to optimize the classifier based on BP neural networks.Through the PET/CT image processing,the functional characteristics,structural characteristics and clinical information of the lesions are extracted as input samples of the neural network based classifier.Then,the benign and malignant diagnosis of the solitary pulmonary nodules is realized by the novel classifier.Classify experimental results on a large number of experiment data from a hospital and public databases on network show that the optimized algorithm is greatly improved on the classification accuracy,indicating that this method is effective in clini-cal classification of pulmonary nodules.

Key words: Solitary pulmonary nodules,Back propagation neural networks,Genetic algorithm,Classification