Study On Microcalcifications Detection Algorithm In Mammography

With the development of society, the changes of people’s lifestyles, diet habits and the environment are leading to the fatality ratio of breast cancer rising in recent years. Breast cancer has been a health threat to major killer of women. Now mammography is the most effective method for the early detection of breast cancer. It plays an important role in early diagnosis of breast cancer. However, due to the special structure of the breast tissue the information is very limit in breast X-ray image. Especially the microcalcifications which have significant value in early diagnosis of breast cancer may not be able to clearly shown in the image. Even experienced clinical experts also lead to misdiagnosis and missed diagnosis phenomenon. With the rapid progress of computer technology, computer aided detection and identification of MCCs has been a hot and difficult research field. In this paper, some key issues of the computer aided diagnosis technique of breast cancer MCCs are systematically investigated. The most important parts would be described as follows:(1) An adaptive mammograms enhancement method is proposed in this paper. Due to the mammograms are low contrast image, in this paper the breast X-ray image is multi-scale decomposed based on non-subsampled contourlet transform. Then the high frequency coefficients of different scales are enhanced through the strengthen function of design. This method can effectively suppress background and noise while effectively enhance the microcalcifications in the image.(2) Extraction of the region of interest in mammograms. We propose a method of using wavelet analysis and image texture feature extraction to extract the regions of microcalcifications in breast X-ray image. Firstly, the algorithm uses the method of gray level co-occurrence matrix to extract image texture features in the time domain. Then it extracts the statistical properties from the high frequency coefficients of wavelet each layer. Finally, we combine both the characteristics and use support vector machine for training and classification for extracting of the region of interest in mammograms.(3) Calcification detection and system design. Firstly the algorithm finds the seeds which represent the microcalcifications area based on top-hat morphologic operator and wavelet transform. Then it uses region growth algorithm to segment the microcalcifications and achieve precise localization of microcalcifications. Finally,we use the matlab software to design a practical microcalcifications detection system to achieve the detection visualize. The system can help the doctor for early breast cancer detection and greatly improve the efficiency of work.The result which compares with the result of subjective detected by doctor shows sthat the algorithm of the breast microcalcifications detection in this paper is effective and practical. It provides an effective method for diagnosis of early breast cancer for clinicians by using computer aided diagnosis system.