Classification of GRF texture in mammograms through discriminant analysis

Modern mammography is the only technique that has demonstrated the ability to detect breast cancer at an early stage and with high sensitivity and specificity. The search for features in this kind of image is complicated by the higher-frequency textural variations in image intensity. The interpretation of mammograms is a skilled and difficult task. This paper deals with the problem of unsupervised classification images modeled by Gibbs Markov Random Fields (GMRF) where the model parameters are unknown. Our approach consists of investigating the variation of the parameters estimations in the multilevel logistic (MLL) model using discriminant analysis (DA). Results obtained from the experiments, show that DA turned out to be an excellent classifier of areas with and without calcifications. Hence the MLL along with DA has been demonstrated to be an effective aid to the radiologist for classifying textures whether containing calcification or not.