Using learning classification and four-dimensional parametric modeling for the analysis of myocardial thickening

A previously, presented 4-D method for modeling the myocardial surfaces and their deformation was refined and applied to the measurement of diagnostic parameters. The method initially defines the myocardial surfaces of the left ventricle. Based on the derived model, measurements of myocardial thickness and thickening in time were produced and used to construct 3-D myocardial thickness maps, which were color coded on the surface for visualization over time. Estimations of myocardial strain maps were also produced, taking into account the deformation of myocardial surfaces. The shape extraction method was improved by utilizing a learning segmentation process, based on a generating-shrinking neural network classifier. A multiscale approach was also adopted which starts from a rough approximation of the expected shape and gradually proceeds to the accurate model. The method was applied to multi-slice multi-phase MRI cardiac acquisitions. Although the displacement and strain maps were not derived from true functional data, have shown promise for cardiac function diagnosis