Creating Segmentation Masks for Benchmark in Digital Mammography

Abstract

Computer aided diagnosis (CAD) as a fast-developing area in medical practice relies on a good preprocessing of images. There are two general image acquisition technologies, analog or film based and digital. To be able to use analog images in CAD applications it is necessary to digitize them and preprocess them so satisfy certain standards. Preprocessing steps usually include intensity equalization and segmentation of objects of interest from the background. In this paper a methodology for automatic mask extraction from manually segmented mammograms is proposed. Medical imaging generally relies on accurate segmentation for CAD applications and it is necessary to have a good ground truth images to benchmark the performance of a given segmentation method. The proposed method describes the entire process of mask extraction using both printed and digitized images. In medio-lateral oblique (MLO) images there are certain key-points which need to be properly detected and segmentation needs to be made according to them. Image alignment process and extraction of the breast tissue and the pectoral muscle from each mammogram available in the mini-MIAS database is proposed.