Neural networks for the classification of benign and malignant patterns in digital mammograms

This chapter presents neural network-based techniques for the classification of microcalcification patterns in digital mammograms. Artificial Neural Network (ANN) applications in digital mammography are mainly focused on feature extraction, feature selection and classification of microcalcification patterns into ‘benign’ and ‘malignant’. An extensive review of neural techniques in digital mammography is presented. Recent developments such as autoassociators and evolutionary neural networks for feature extraction and selection are presented. Experimental results using ANN techniques on a benchmark database are described and analyzed. Finally, a comparison of various neural network-based techniques is presented.