Texture analysis of ultrasonic liver images based on spatial domain methods

The paper introduces three texture analysis methods of ultrasonic images based on spatial domain method. Feature parameters, including mean, variance, contrast, homogeneity, angular second moment and entropy, are achieved from gray histogram statistic, gray level difference statistic (GLDS), gray level co-occurrence matrix (GLCM). Then the above statistical feature parameters are applied for texture classification by neural network. The Probabilistic Neural Network (PNN) is employed as a classifier to differentiate ultrasonic fatty liver image from normal liver image. Experimental results showed that the joint statistical feature parameters extracted from the three methods achieve good effects.