Segmentation of B-mode cardiac ultrasound data by Bayesian Probability Maps

摘要: In this paper we present a model for describing the position distribution of endocardium in two-chamber apical long-axis view heart clinical B-mode ultrasound cycles. We propose novel Bayesian formulation, including priors spatial and temporal smoothness, preferred shapes position. The shape takes into account both endocardium, atrial region apex. likelihood is built using statistical signal model, which attempts to closely censored signal. addition, use Gamma mixture with unknown censoring point, handle artefacts resulting from left-censoring US B-mode, our knowledge novel. posterior density sampled by Gibbs method estimate expected latent variable representation call Probability Map; map describes probability pixels being classified as within endocardium. regularization parameters are estimated cross-validation, results compared against Chen et al.