The Research Of Mammograms Computer-aid Diagnosis

Breast cancer is one of the most common malignant diseases. The incidence ofbreast cancer has been increased quickly in recent years. In2011, there were230480women get breast cancer in the United States. The incidence of breast cancer is thehighest among female. In China, statistical data in Beijing, Tianjin and Shanghai showthat the breast cancer is also a common cancer, and the incidence rate is increasing.Breast imaging technologies are effective methods for breast cancer diagnosis.The commonly methods include mammography, MRI and B ultrasonic. Among them,mammography is the most common method. To effectively process the huge imagedata, computer-aid detection (CAD) is an important research subject, such as tumordetection, calcification detection and the classification of breast density. There isresearch show that there exists strong relationship between the risk of breast cancerand breast density. So breast density evaluation has important clinic application value.This paper discusses the computer aided breast density estimation. It includesfollowing contents:(1) Show the method of classification of breast density based on local region.Compared to traditional methods which are based on global breast region, this methodis based on local region. In this method, the breast region is divided into rectanglelocal regions, and then the SVM classify local regions into dense region and sparseregion. By calculating the ratio of dense region, we can get breast density. Theexperiment results proved that this method can achieve better effect for breast densityclassification than traditional methods.(2) For the defect of the method of classification of breast density based on localregion, the way of image preprocessing is improved. In this method, the preprocessingis based on wavelet. Through experimental observation, we find that guard bands ofgland region and fat region are different. So we process the coefficients of waveletdecomposition in appropriate way. And after wavelet reconstruction, gland region willbe enhanced. The experiment results proved that effect of this method is better thanthe first method, especially for the image of which breast density is lower25%.