Features selection for detection of tumors on mammograms

摘要

Selection of input features is a key stage in building a classification system. In this paper, we focus on the problem of features selection of a CAD system for detection of breast tumors on mammograms. Here, 29 features were calculated for each of the cancerous candidates. Most of these features have been used in detection of malignant masses in digitized mammograms. We use the forward stepwise selection method to select an optimum feature subset for our CAD system. The effectiveness of the selected features is demonstrated on 208 X-ray mammograms by comparing with our previous work