Genetic programming and feature selection for classification of breast masses in mammograms

A dataset of 57 breast mass mammographic images, each with 22 features computed, was used in this investigation. The extracted features relate to edge-sharpness, shape, and texture. The novelty of this paper is the adaptation and application of genetic programming (GP). To refine the pool of features available to the GP classifier, we used five feature-selection methods, including three statistical measures -- Student's t-test, Kolmogorov-Smirnov Test, and Kullback-Leibler Divergence. Both the training and test accuracies obtained were above 99.5% for training and typically above 98% for testing.