Research Of The Breast Tumor Diagnosis Method Based On Neural Networks

Breast tumors are divided into benign breast tumor and malignant breast tumor. Malignant tumor of the breast is ranked first in the malignant tumor which causes great harm to women’s physical health, and it has been developed fastly in recent years in the major cities in China. The scientific breast tumor diagnosisi method not only can detect the benign breast tumor and take measures to prevent its deterioration, but also to prevent the misdiagnosis of malignant breast tumor which will delay the time of treatment. Therefore, It is very necessary to use scientific method to achieve the effective diagnosis of breast tumor. A large number of high quality samples of breast tumor cases which are provided by medical experts make the building of scientific breast tumor diagnosis model possible.Because of artificial neural network’s powerful learning ability, non-linear mapping ability and the excellent characteristics of associating, generalization, analogy and generalization,it is widely used in medical diagnosis whose inner mechanisms are very difficult and complex.So it is appropriate to use neural network to set up different scientific breast tumor diagnosis method models and find the most available breast tumor diagnosis method model based on different neural networks by comparisons.The main contents of this dissertation are as follows:(1) Research on breast tumor diagnosis method based on BP neural network.BP neural network is the most widely used neural network with clear principle and simple structure.But there are many parameters needed to set up in the BP model,and these parameters have an important influence on the convergence speed, the error, and the accuracy. The research of the regulations of how to set up BP neural network parameters will provide reference for constructing superior performance model.(2)Research on optiamizing inputs by Genetic algorithm.In BP model,too many inputs will degrade the performance of the model. With selecting important inputs and removing redundant inputs, it not only degrade the performance of the model,but also reduce the time of modeling.GA is a very effective global optimization algorithm. Through the research of the input variables coding method, the setting of evaluation function and the parameters.it will select important and proper inputs.These important and proper inputs can optimize the breast tumor diagnosis method model based on BP neural network.(3)Research on breast tumor diagnosis method based on ANFIS.ANFIS not only has strong learing ability and nonlinear mapping ability as same as neural network,but also has fuzzy information processing capability as same as fuzzy theory.Therefore,it is a very excellent kind of intergrated neural netwok. The fuzzy information processing capability of ANFIS can well deal with the breast tumor diagnosis samples. The research of the breast tumor diagnosis model based on ANFIS will provide another solution of the breast tumor diagnosis.(4)Research on comparisons between breast tumor diagnosis method models based on different neural networks.Breast tumor diagnosis method models based on different neural networks are compared from the process of modeling and the results of the experiments.The reasons why different models have different performances are pointed.With the analysis.it can provide reference for the building of breast tumor diagnosis model.At the end of this dissertation,the main research is summarized.It not only points out the main research results,also points out the further researches.