This project studies heart diagnosis tools based on 3D ultrasound techniques. The project has three main goals: i) the development of image reconstruction and heart measurement algorithms for the analysis of the cardiac cycle and computation of clinical parameters (ventricular volume, ejection fraction and wall thickness); ii) implementation of an experimental set up for the acquisition of 3D data during medical examinations of the heart and iii) clinical evaluation of the 3D ultrasound algorithms developed in the project Bayesian reconstruction methods will be used to estimate a 3D+T model of the heart at different instants of the cardiac cycle. The region of interest will be described using a multi-scale representation based on 3D splines. The motion and deformation of the heart cavities will be obtained by segmenting the reconstructed volume at each instant of time. To fill the gaps between the inspection planes some kind of interpolation has to be devised. This operation is embodied in the Bayesian reconstruction provided that an adequate prior is used. Unfortunately popular Gaussian priors have an undesirable smoothing effect at the boundaries, which degrades the estimation of the heart walls. Therefore, a discontinuity preserving prior will be used instead. Another key aspect for achieving high quality results concerns the data model used for reconstruction.