Statistical modeling of B-Mode clinical kidney images

Envelope of B-Mode ultrasound is generally modeled by using Rayleigh, Rician, K, Nakagami (Generalized), Weibull, Gamma (Generalized), Lognormal, Normal and other distributions. Estimation of parameters is done using the method of moments or through Maximum Likelihood Estimation. The paper proposes a mathematical model of ultrasound kidney images of at different stages of growth. Images are obtained at different time from a commercial ultrasound machines in clinical settings. Using Nakagami distribution and Generalized Gamma Distribution (GGD) these images are modeled. The parameters of employed distributions are estimated using MLE. Based on estimated parameters the Nakagami and Generalized Gamma Distribution (GGD) are fitted to the empirical histogram corresponding to Ultrasound B-mode images. Statistical characteristics of clinical ultrasound B-mode images were done for classification of the central and peripheral region in kidney. The efficacies of both the distributions are evaluated in terms of Kullback-Leibler (KL) measure. The results indicate that classification based on GGD to be better.