A comparison of the Homodyned K-distribution and the single distributions for RF ultrasonic speckle modeling

Abstract

For observing the parameters and the fitting performance, this paper compares the Homodyned K-distribution with the single distributions for RF ultrasonic speckle modeling. To implement different scatterer distributions representing a variable density of random scatterers with or without coherent component, A set of 3D scatterer models are built based on a three-dimensional Hilbert curve following Gamma distributions with different values of shape and scale parameters. The RF data are simulated by using the Field II software. Then the maximum likelihood estimation (MLE) for statistical histograms of the energy of the RF data is performed to obtain the values of log-likelihood and model parameters. In order to evaluate the fitting performance and parameter meaning of the HK distribution, the mean and standard deviation of these estimated values are compared with those based on the optimal fitting model chosen from commonly used single-distributions (OSD), the K, Rayleigh and Rician distributions. The results show the parameters of Homodyned K-distribution obtained by the MLE could independently represent the clustered, random or uniform characteristics for scatterer distribution. However, the fitting accuracy could only catch up with that based on the OSD joint model under the condition that the tissue contains the scatterers from medium to high effective-density, as well as deterministic or non-deterministic components. The OSD model is still a better choice in the case of the fitting performance emphasized in practice, especially the tissue with a wider range of scatterer densities and deterministic components.