Quantification of three-dimensional left ventricular segmental wall motion and volumes from gated tomographic...

Tomographic radionuclide ventriculograms may be used for three-dimensional wall motion analysis. We propose that automatic quantification of these images is possible, and here we describe the implementation and validation of a method to perform this task. Automatic computer methods were developed to locate the left ventricular (LV) endocardial surfaces in all time frames of the cardiac cycle. Global, regional, and local motion and volume were computed. Results were displayed using three-dimensional graphics. The methods were validated using phantom, canine, and human studies. Actual phantom values correlated well with experimentally determined volumes, y = 1.01x + 1.29ml, r = 0.99. In the canine model, the LV endocardial surfaces were located to within an average of 1.9 mm and 3.7 mm at end-diastole and end-systole, respectively. Areas of obvious wall motion abnormalities in automatically processed patient studies corresponded well with angiographically documented coronary artery disease. End-diastolic and end-systolic volumes computed automatically from single photon emission computed tomography averaged errors of 9% and 38%, respectively, when compared with contrast ventriculographic volumes. These results indicate that it is possible to automatically identify the left ventricular endocardial surface in gated tomographic radionuclide ventriculograms. The location of these surfaces corresponds well with the location of implanted endocardial markers, and global volume computed from these surfaces corresponds well with known volumes.