Diagnosis Of Benign And Malignant Breast Tumors Using Multi-source Data Fusion

The proportion of breast cancer in women's cancer is growing.Breast cancer is a disease with hidden symptoms.It is not easy to be found in clinical detection.Many people have become advanced cancer when they are diagnosed with breast cancer by a doctor.In order to effectively reduce the mortality of breast cancer,we urgently need to increase the publicity of early detection and treatment.Malignant tumor is an important manifestation of breast cancer.Computer-aided diagnosis is of great significance for the automatic recognition of benign and malignant breast tumors.This thesis is different from the previous research.Based on the analysis of the characteristics of breast tumors,a new method of fusion imaging features and medical clinical features is proposed.Then,in view of the new features produced by the fusion,we propose a unique classification method and an improved classification algorithm for benign and malignant tumors.In this paper,the specific research contents are as follows:Firstly,we analyze the specialties of multi-source breast tumor data and propose a fusion method of image features and clinical features,namely,semantic feature group fusion algorithm based on canonical correlation analysis.The algorithm is used to analyze the region of interest in breast cancer and produce qualitative semantic features.Then we group the high-level semantic features and set the feature fusion rules.The feature group fusion algorithm based on canonical correlation analysis is applied to fuse the image features and clinical features,so as to generate a new breast tumor feature,namely the fused feature.The fusion features are more beneficial to the benign and malignant classification of breast tumors.Secondly,the problems and difficulties in the classification of fusion features are discussed around the inherent characteristics of the fused features.Then we introduce the idea of kernel function and weighted integration and propose a weighted ensemble classification method based on fusion features.Experimental results show that the ensemble classification method not only has better classification accuracy and generalization ability,but also achieves remarkable results in solving the two classification problem of fusion features.Thirdly,aiming at the classification of benign and malignant breast tumors,an integrated classification method combining Random Subspace and Ada Boost ELM is proposed.This method can improve the deficiency of ELM in the classification of breast tumors and improve the performance of integrated classification.The experimental results show that the prediction error of this method is lower than that of ELM and other mainstream integrated learning methods.