The Research On Computer Aided Detection Of Microcalcifications In Mammogram

Breast cancer is the top cancer in women both in the developed and the developing countries. The disease accounted for 0.5 million deaths every year in the world. The number of deaths is expected to continue rising, with 2%-3% rate annually. Early detection in order to improve breast cancer outcome and survival remains the cornerstone of breast cancer control. Mammography screening is a technique for early breast cancer detection and is the only screening method that has been proven to be effective.Microcalcifications are the important early signs of breast cancer which can be de-tected by mammography. Because of the nature properties of microcalcification, subtle size (0.05mm-1mm), low contrast in high density, varying shapes and blurred margin, the detection work becomes difficult to radiologist.Computer aided detection (CADe) is developed to help radiologist to improve the detection precise. It can automatically mark the suspicious regions with microcalcifications by the help of computers. However, the sensitivity and specificity of the CADe are still not satisfactory up to now.In this thesis, we studied the algorithms in all CADe steps. By improving the process-ing precise in each CADe step, we can improve the sensitivity and specificity of CADe. The researches include the segmentation of the breast region in mammogram, the segmen-tation of the pectoral muscle in mammogram, intensities equalization of mammogram, and microcalcifications detection in mammogram. The innovations include:(1) Proposed a pulse coupled neural network based breast region segmentation algorithm to extract complete breast region from a mammogram with noisy background.(2) Proposed a discrete time Markov chain and active contour model based pectoral mus-cle segmentation algorithm to extract pectoral muscle from a mediolateral oblique view mammogram. (3) Proposed a novel intensities equalization method, which requires only the statistic information of a mammogram before equalizing its intensities.(4) Improved fast marching based multi-objects segmentation algorithm and reduced its time complexity from O(N log2 N) to O(N).(5) Proposed a blob based microcalcification detection algorithm to mark the microcal-cifications from a mammogram accurately.The innovations not only improved the processing precise in each CADe step, but also broadened the research ideas of CADe. Some algorithms, such as discrete time Markov chain based boundary detection algorithm, fast marching based multi-objects segmentation algorithm, etc., can also be applied to other more general image processing fields.