The aim of this work is to track specific anatomical structures in temporal sequences of echocardiographic images. Ultrasound images are available in two broad data types: raw or video data. Different stochastic processes using different kind of information are compared on the basis of these two data types. We explain the selection of a particular model w.r.t. the type of data, and describe the relevant properties that must be taken into account to obtain the best possible results. The models are expressed within a Markov random field framework and we also discuss parameter estimation and energy minimization for the different models.