A model-based procedure for fully automated boundary detection and 3D reconstruction from 2D echocardiograms

Describes an automated procedure for boundary detection and 3D reconstruction in echocardiography. Boundary detection is carried out by a combination of knowledge based and dynamic programming procedures. The knowledge based procedure includes a mechanism which controls and evaluates a set of image analysis procedures. The dynamic programming procedure is based on the minimization of a cost function including terms representing echo intensity, edge strength, and boundary continuity in space and time. The 3D reconstruction procedure is based on extracted boundaries from two short-axis and three apical views in conjunction with a geometric model. By replacing conventional bi-plane Simpson method for stroke volume estimation with the authors´ model-based scheme, correlation between echocardiographic and dye dilution based estimates increased from 0.69 to 0.88.<>