Cylindrical raster map-based surface reconstruction of the left ventricle using real-time three-dimensional ultrasound

An important part of diagnosing heart disease is assessing the mechanical function of the heart. Ejection fraction of the left ventricle is one of the main indicators of the heart's systolic function. It can be measured using cineangiography, which uses geometric assumptions in calculating volume, or tomographic imaging systems such as CT, MRI, and 2D ultrasound, which require cardiac and respiratory gating. A new method of extracting the left ventricular endocardial surface from real-time 3D (RT3D) ultrasound data is presented in this dissertation. RT3D ultrasound is a volumetric imaging system that does not require gating and therefore can make beat to beat measurements of the heart. The basic design of the new method is to use tracings of the endocardial border on a set of arbitrary slices through the 3D ultrasound volume and reconstruct the surface interactively using geometric characteristics of the left ventricle to help make the process faster. Analysis of the surface reconstruction process showed that only a 25% tracing was needed for a good reconstruction of the left ventricle. The study of the border localization error revealed that there is a consistent underestimation of the left ventricular volume when tracings are done at the perceived border position. The experimental results showed that the new method could accurately measure volumes with only a partial tracing of the left ventricular surface. Overall, the new method performed almost exactly as predicted and was faster and more accurate than manual tracing.