Joint Segmentation and Deconvolution of Ultrasound Images Using a Hierarchical Bayesian Model based on Generalized Gaussian Priors

Ningning Zhao, Adrian Basarab, Denis Kouame, Jean-Yves Tourneret

Published 2014-12-08Version 1

This paper addresses the problem of ultrasound image deconvolution within a Bayesian framework. Ultrasound images exhibit characteristic speckle patterns previously shown to be strictly correlated with the tissue structures. As a result, the statistical properties of the speckle have been extensively studied as source of information for applications such as image segmentation, deconvolution and tissue characterization. We investigate herein the generalized Gaussian distribution which has been recently shown to be one of the most relevant distributions for ultrasound radio-frequency signals. By introducing the labels for the image pixels, we introduce a new Bayesian model allowing the ultrasound images can be jointly segmented into several statistically homogeneous regions and deconvolved to provide the tissue reflectivity map. The Bayesian estimators associated with the proposed model are difficult to be expressed in closed form. Thus, we investigate a Markov chain Monte Carlo method which is used to generate samples distributed according to the posterior of interest. These generated samples are finally used to compute the Bayesian estimators of the unknown parameters. The performance of the proposed Bayesian model is compared with existing approaches via several experiments conducted on realistic synthetic data and in vivo ultrasound images.