Comparative Evaluation of Statistical Pattern Recognition Techniques for the Classification of Breast Lesions

In this study we have investigated shape features, extracted from the microcalcifications, constrast and texture features, extracted from the region of interest (ROI), to classify early breast cancers which has microcalcifications associated. These features were analyzed using three statistical classifiers: two Bayesian classifiers — linear classifier (LC), quadratic classifier (QC) — and the K-Nearest Neighbor method (K-NN).