Left ventricular shape reconstruction and description in 3D with application to self-referring echocardiographic reports

In order to improve the evaluation of echocardiographic stress tests, 3D reconstruction of left ventricular (LV) endocardial surface with a complete description of the shape and its modifications was obtained using a limited number (4-6) of long-axis apical views. Regularization theory was exploited in order to minimize an energy functional, i.e. the weighted sum of fitting mean square error and thin-plate surface variation. According to the finite element method, energy minimization was accomplished iteratively by means of the conjugate gradient descent algorithm and considering either a cylindrical parametrization or a spherical one. Koenderink's index was computed to map the endocardial surface onto a matrix of compact and invariant shape descriptors. Grouping and matching of hyperbolic and elliptical patches allowed regional deformation due to acute ischemia to be identified, whereas little variation was found in normal subjects.