Abstract

The Nonlinear principal component analysis (NLPCA) based image classification on mammogram data is proposed in this paper. A feature extraction is developed on the basis of NLPCA which is a nonlinear method. Multilayer neural network normally implements nonlinear components. The relationship between the variables is not linear in complex problems where the NLPCA is needed to achieve better performance in contrast to the linear PCA. Normal and abnormal are the two classes of NLPCA technique which is used to classify each segment. Only normal patterns are used during the algorithm training phase and for segmentation we use two nonlinear features in classification. The recurrent neural network (RNN) is used to model the distribution of these features. Results are more accurate which is shown by the classification obtained using NLPCA. KNN classification is applied on nonlinear components for the evaluation of further enhancements. The proposed method enhances the classification performance through the reduction of dimensionality and optimal component set which is shown in the simulation.

Paper Details