Active Shape Models and Segmentation of the Left Ventricle in Echocardiography

摘要

Segmentation of the left ventricle in echocardiographic images is a task with important diagnostic power. We propose a model-based approach that aims at extracting the left ventricle for each frame of the cardiac cycle. Our approach exhibits several novel elements. Modelling consists of two separate components, one for the systolic and one for the diastolic moment. Segmentation is considered in two steps. During the first step a linear combination of the sy stolic and the diastolic model is to be recovered - that dictates the new model - along with a similarity transformation that projects this model to the desired image features. During the second step, a linear combination of the modes of variation for the systolic and diastolic models is recovered for precise extraction of the endocardium boundaries. The process is considered in the temporal domain where constraints are introduced to couple information across frames and to lead to a smooth solution. Promising results demonstrate the potentials of the presented framework.