The Application Of Fuzzy Neural Network Optimized By Genetic Algorithm To Lung Cancer Diagnosis

In the world today, Lung cancer has become the most common malignancy. In China, the incidence of lung cancer is rising yearly, and the death number of lung cancer is the biggest now. The probability of lung cancer for men is bigger than women, and the incidence in urban is larger than that in rural palace. There are few obvious symptoms or some symptoms are same with other disease, so it is hard to detect lung cancer in early period, and sometimes when we find it has been the end stage. The earlier we detect it, the easier we defeat it, so there is very important realistic significance to improve the method of early detection for lung cancer, which can increase the survival rate and save more people.In this paper. I make an introduction to the history of lung cancer detection. Artificial neural network and fuzzy neural network was talked about. Fuzzy neural network comes from the combination of artificial neural network and fuzzy theory, which has the advantage of both them. There are strong learning ability, self-adaptability and self-organization. Dealing with fuzzy information with neural network can solve the problem of extracting fuzzy rules and making membership functions. Fuzzy technology has great critical thinking and reasoning abilities, which can improve the ability of dealing with fuzzy information for neural network. But there are ineluctable drawbacks for neural network, such as the low convergence rate and relative minimum. Genetic Algorithm comes from the law of natural development-biological evolution, which includes natural selection and biology genetics. It uses inheritance, variation and some other steps to add the fitness of individuals. The combination of neural network and genetic algorithm can optimize the weight and threshold of neural network, which can produce a better structure of neural network, so the network can have good predictive accuracy.Methods:The genetic algorithm is used to optimize the fuzzy neural network to detect lung cancer.13 binary parameters, which are fuzzed with Gaussian membership function, and 13 non-binary parameters are used as net input. Firstly, a simple BP neural network is built; Secondly, optimize the weights and thresholds; Next, train the training sample with the new network which is optimized by the genetic algorithm; Then, test the network with the test set. In the end, make a contrast with the original fuzzy neural network.Conclusions:Fuzzy neural network which is optimized by genetic algorithm has a better performance in prediction of probability of malignancy in pulmonary nodules than original neural network. There are 1 false-positive and 1 false-negative in new network. But in original network, there are 2 false-positive and 2 false-negative. The diagnostic accuracy rate of new network is about 96%, which is 4 percentage points higher than that of original network. So optimizing neural network with genetic algorithm and using it in lung cancer detection have great value.