RECOVERY OF THE 3-D SHAPE OF THE LEFT-VENTRICLE FROM ECHOCARDIOGRAPHIC IMAGES

A computational method is reported which allows the fully automated re covery of the three-dimensional shape of the cardiac left ventricle fr om a reduced set of apical echo views, Two typically ill-posed problem s have been faced: 1) the detection of the left ventricle contours in each view, and 2) the integration of the detected contour points (whic h form a sparse and partially inconsistent data set) into a single sur face representation, Our solution to these problems is based on a care ful integration of standard computer vision algorithms with neural net works. Boundary detection comprises three steps: edge detection, edge grouping, and edge classification, The first and second steps (which a re typical early-vision tasks not involving specific domain-knowledge) have been performed through fast, well-established algorithms of comp uter vision, The higher level task of left ventricle-edge discriminati on, which involves the exploitation of specific knowledge about the le ft ventricle silhouette, has been performed by feedforward neural netw orks, Following the most recent results in the field of computer visio n, the first step in solving the problem of recovering the ventricle s urface has been the adoption of a physically inspired model of it, Bas ically, we have modeled the left ventricle surface as a closed, thin, elastic surface and the data as a set of radial springs acting on it, The recovery process is equivalent to the settling of the surface-plus -springs system into a stable configuration of minimum potential energ y, The finite element discretization of this model leads directly to a n analog neural-network implementation, The efficiency of such an impl ementation has been remarkably enhanced through a learning algorithm w hich embeds specific knowledge about the shape of the left ventricle i n the network, Experiments using clinical echographic sequences are de scribed, Four apical views (each with a different rotation of the prob e) have been acquired during a heartbeat from a set of seven normal su bjects, These images have been utilized to set the various processing modules and test their capabilities.