Mammography Enhancement And Detection Based On Directionlet And Sparse Representation

Breast cancer is one of the most common and high-risk malignant tumor diseasesamong women, seriously affecting the women’s health and even threatening their life.The breast cancer CAD (Computer Aided Detection, CAD) system consists of two parts:mammography enhancement and detection. Mammography enhancement couldeffectively enhance the contrast of lesion area and suppress the noise, providing morereliable diagnosis data for doctor. Mammography detection firstly study the pathologicfeatures and use sliding window to detect the mass area, then label the suspicious areas,which can assist the doctor to diagnose the disease more easily and exactly.This paper delves into the theory of directionlet transform and sparserepresentation, and then applies them into the enhancement and detection algorithm.The main contents of the thesis are summarized as follows: the directionletdecomposition is achieved by designing of undecimated scheme. Considering thecorrelation of the frequency coefficients, a new image denoising and enhancingalgorithm is proposed. It is shown that the enhancing algorithm can suppress the noiseand enhance the edge of image. Secondly, build the feature dictionary throughextracting the gray feature and HOG feature from the ROI, and then use the slidingwindow and sparse classifier to achieve the detection results of abnormities.Experimental results show that the proposed enhancement method can enhance thecontrast of lesion area and the proposed detection method can quickly and effectivelytrack the mass area.