A fully automated scheme for mass detection and segmentation in mammograms

A novel automated mass detection algorithm is presented in this paper. There are two main steps in this method. First it establishes a sech template to simulate the masses and employs template matching to obtain an image which measures the suspicious degree of every pixel in the segmented breast. An adaptive thresholding technique based on maximum entropy principle is used on the exponential transformed feature image to get some regions of interest (ROIs). Then region growing is applied on the background corrected regions of interest to separate the masses from the background. This method has been tested on the Digital Database for Screening Mammography (DDSM). Preliminary results display a best sensitivity of 97.2% with 1.83 false positives per image.