Texture Analysis of Ultrasonic Liver Image Based on Wavelet Transform and Probabilistic Neural Network

In this paper, a new liver texture classification method based on discrete wavelet transform (DWT) is proposed to differenciate ultrasonic normal liver image from fatty liver. The low frequency coefficients of transformed image denote the approximation of the image, and the high frequency coefficients denote the details of the image. So the character deduced from DWT coefficients can embody texture feature of images. The mean and standard deviation of transformed sub-image are extracted. Neural networks are employed to classify pattern based on learning from examples. The probabilistic neural network (PNN) develops from radial-basis- function neural network (RBFNN) and approaches the Bayes theorem maximum posterior probability, thus it is employed as a classifier in our method. Then the above statistical features are applied for texture classification by PNN. Experimental results show that statistical features extracted from DWT coefficients achieve good effects.