Breast and Pectoral Muscle Contours Detection Based on Goodness of Fit Measure

Mammography is one of the primary breast imaging methods. A mammogram is a 2D projection of the compressed breast. Computer-aided methods have been developed for automatic analysis. In this paper, a novel approach based on statistical theory is proposed. The concept of ´goodness of fit´ in statistics theory is utilized to develop a measure of the local spatial distribution in mammogram. The method is implemented on mammograms pixel by pixel. The background and the pectoral muscle can be separated by detecting their contours on the resulting images. The proposed method has been applied on images in mini-MIAS database.