改进的Adaboost集成神经网络技术在乳腺疾病辅助诊断模型中的应用-计算机应用技术专业论文.docxVIP

II II Abstract With the increasing incidence of breast cancer found in women worldwide, the outpatient workload of breast surgery is on the increase. It is meaningful to research and develop a Aided Diagnosis Expert System which can not only help doctors diagnose disease to improve the clinic work of breast surgery efficiency, reduce or avoid missed diagnosis, misdiagnosis, but also aid young intern doctors to gain medical experiences at the same time as well as allow women to test breast disease on their own. However, when we analyzed the Medical Aided Diagnosis Expert System developed in the past, we found that the Medical Aided Diagnosis Expert System can not simulate the diagnosis process of a doctor because of the defects in the Medical Aided Diagnosis Expert System and the presumed diagnosis effect of the system is not ideal. The paper tries to lead in the integrated neural network technology in the process of constructing the aided diagnosis model aimed at this defect. As long as the results of individual neural network prediction results are better than random guess by doing experiences, the prediction effects of integrated neural network is certainly superior than the individual neural network. We also found that the cost of the integrated neural networks in classification error of the sample is not the same, and the cost of benign lesions diagnosed as malignant tumors is repetitive inspection which is less harmful; But cases of malignant tumors misdiagnosed as benign lesions will adversely affected the chances of treatment which is more harmful. In this case, we have to improve the algorithm of traditional integrated neural network-Adaboost, by changing the initial weights in the algorithm, and dynamically updating weights and the threshold, the integrated neural network pays more attention to the classification of subclass. Having changed the traditional concept based on the minimum error rate and constructed the Aided Diagnosis Expert System on the basis