Automatic outlining of the left ventricular cavity in magnetic resonance imaging. Comparison with manual outlining

The purpose of this work was to build an automatic outlining algorithm of the left ventricular cavity. This algorithm is based on Mathematical Morphology. The main tool of segmentation is nammed Watersheds and Geodesic Reconstruction algorithms. For this study, we Implemented an uptodate version of the algorithm by means of Hiearchic FIFO. On the native magnetic resonance image, we studied the reproducibility of human manual tracking. We obtained correct automatic outlines of the left ventricule for 21 prospectively examined subjects. This study demonstrates the advantage of an automatic approach according to the important variability in manual tracing and to the big number of images to study.