An Automated System for Detection and Segmentation of Masses in Digital Mammograms

摘要

A computer scheme capable of providing objective information may aid radiologists in their diagnosis of masses, thus preventing unnecessary biopsies. In this paper, a computer-aided system for automatic detection and segmentation of masses in mammograms is presented. The system is designed based on a sequence of successive preprocessing procedures followed by the PNN detection and entropic thresholding-based segmentation methods. The preprocessing is carried out by two stages. The first stage enhances the gradients of pixels with high intensities, and the second stage uses a median filter to eliminate the noise effects in the background of ROI. In order to do the detection and segmentation of masses, five texture features are used for differentiating masses from normal breast tissue on mammograms. The approach of detection and segmentation involves (1) automatic masses detection using texture features on PNN and (2) automatic masses extraction using entropic thresholding-based method. The designed system was tested by a data set provided by Taichung Veteran General Hospital, the results showing promises of the proposed system.