Preprocessing And Detection Of Breast Arterial Calcification In Mammograms

Nowadays, breast cancer has become the most common threat to the health of women. Europe, America and Australia have been the areas of high incidence of breast cancer. Asia is low relatively. But data shows that the breast cancer and the death of breast cancer is rising world widely every year. Therefore, early detection, early prevention and early treatment are key factors in reducing breast cancer mortality rate. Although many diagnostic methods in recent years, but the X ray mammography is still the most common way of checking the disease. However, due to image quality, technicians and doctors own relationship factors, it is likely to cause misdiagnosis and missed diagnosis. Therefore, the mammo CAD (computer-aided diagnosis) system based on digital X-ray images is produced to help medical examination and diagnosis.Researches related in the paper is based on Mammogram CAD, including the pre-processing and Breast Arterial Calcification detecting algorithm in the micro calcification detection. Micro-calcifications are small high points in the mammograms. Micro-calcification cluster is a sign of early breast cancer so that detection of micro-calcification is important in the early time for preventing breast cancer. For a long time, researchers spent large engorge on micro-calcification detecting, which makes detecting rate of micro-calcifications is satisfying. However, high detecting rate also rises the false positive rate. In this paper, we propose a new pre-processing and false positive decreasing algorithm. In the pre-processing part, we compare two methods. One of them is the normal method that using the watershed method to find the breast skin-line after double thresholding the image. The other method is based on K clustering to find the skin-line according to the features that the pixel vales decrease along the way from inside to the skin-line. Then we do the expolating and find the accurate skin-line. The breast arterial method is based on the Hessian matrix and multi-scale analysis. First of all, we calculate the Hessian matrix of every pixel and find their eigenvalues. Using the eigenvalues, we can get a vesselness map with a vesselness formula. After that, we threshold the image and determine the initial breast arterial calcification areas. Then we combine morphology,labeling and features such as roundness to drop the normal calcification and Coarse calcification and find the accurate breast arterial calcification areas.