Research And Implementation Of Breast Tumors Diagnosis Based On Density Distribution

According to relevant statistical data,the number of female who died for the reason of breast cancer increased,which has caused serious damage on the physiology and psychology of female.If breast lesions can be found earlier and appropriate medical treatment can be taken,the mortality will be reduced.In clinical,mammography has become a common mean for early detection of breast cancer.To avoid the shortcomings of subjective and high misdiagnosis rate caused by traditional reading,computer aided diagnosis(CAD)system emerges.The computer aided diagnosis system mainly includes the detection of breast tumors and diagnosis of benign and malignant tumors.CAD,as the second reference,can provide objective and effective information for radiologists to improve the diagnosis rate of breast cancer and to reduce the false negative rate.The main contributions are listed below:(1)Automatic detection of breast tumors.There is high connection between breast density and breast cancer,so breast density is an important index to predict the risk of breast cancer.In this paper,the segmentation method of tumors is density clustering based on sub blocks,the breast region is divided into some sub blocks without overlapping;local density features of breast are extracted for each sub block,and density feature vector model is established;feature vectors are reduced and clustered by using US-ELM and SP-FCM algorithms,tumor region with dense density is clustered,and normal tissues are discrete distribution.Finally,the results of breast tumors segmentation and detection are obtained.(2)Diagnosis of benign and malignant breast tumors.According to the difference between breast tumor density and surrounding normal tissues density,the value of density in the diagnosis of benign and malignant breast tumors was analyzed.In addition to extracting the morphological features and texture features of tumors based on BI-RADS standard,seven density features were extracted and feature vector model was established.Finally,extreme learning machine(ELM)was used as a classifier to diagnose benign and malignant of breast tumors in database.The results showed that density features of breast tumors can be effectively distinguished from benign and malignant tumors compared with other features.In summary,this paper adopted local density features clustering to detect breast tumors,and breast tumors to be diagnosed are analyzed by global density features based on ELM,and the diagnosis of benign and malignant tumors was combined with other features.The important significance of breast density was studied in computer aided diagnosis and detection system of breast tumors.