An information fusion based method for liver classification using texture analysis of ultrasound images

This paper presents a method for classification of liver ultrasound images based on texture analysis. The proposed method uses a set of seven texture features having high discriminative power which can be used by radiologists to classify the liver. Feature extraction is carried out using the following texture models: Spatial Gray Level Co-occurrence Matrix, Gray Level Difference Statistics, First order Statistics, Fourier Power Spectrum, Statistical Feature Matrix, Law’s Texture Energy Measures and Fractal Features. Based upon the results of Linear Discriminative Analysis (LDA) followed by box-plot analysis and Pearson’s correlation coefficient, 7 best features from a set of 35 features are selected. These selected features are then fused using a linear classifier. The novelty of the proposed method is that, it combines the best features from different texture domains along with their weights and ‘weighted z-score’ values. Subsequently, these values are used to compute a discriminative index for liver classification. The results show that this method has overall classification accuracy of 95% and low computational complexity.