Texture Classification for Liver Based on Multifractal and PNN

Abstract

This paper proposes a system applying PNN for classfying the hepatic parenchymal diseases in ultrasonic B-scan texture. In this study,we propose a multifractal method to select the patterns for the training set and validation sets. A modified DBC algorithm is used to calculate the fractal dimensions of the images. FDWT is adopted for feature extraction.The result shows that, compared with the conventional 3-layer NN, the performance of the proposed system is improved.