Feature selection and analysis on mammogram classification
Aijuan Dong, Baoying Wang · 2009
Automatic mammogram analysis is important in early breast cancer detection. In this paper, we present a multi-resolution approach to automated classification of mammograms using Gabor filters. Specifically, Gabor filters of different frequencies and orientations have been used to extract textual patterns of mammograms. To increase classification efficiency and reduce feature space, statistic t-test and its p-values for feature selection and weighting are proposed. Experimental results show that Gabor filters are able to extract textual patterns of mammograms, statistical-based feature selection and weighting can be used to further reduce the feature space without degrading the classification performance.