On the generation of short-axis and radial long-axis slices in thallium-201 myocardial perfusion single-photon emission tomography

We tried to develop fully automatic reorientation algorithms in thallium-201 myocardial perfusion single-photon emission tomography, and tested a method to evaluate the quality of reorientation. The left ventricle was automatically segmented using count density information, contours generated with Laplacian operators in both transaxial and sagittal slices, and morphological and positional characteristics of the contours. Reorientation was automatically performed based on knowledge of the long axis of a second degree surface fitted to the myocardial wall. We tried to achieve improvement in reorientation without relying on any functional description of left ventricular shape. Quality of reorientation was evaluated and improved using interactive tools in combination with radial long-axis slices. Two groups of 50 patients, after stress and rest, were analysed using the traditional manual and the fully automatic procedures. Automatic segmentation was successful in 98 out of 100 cases, and automatic reorientation was of reasonable quality. Reorientation obtained with the radial long-axis slices tool was better than after traditional manual or automatic reorientation. Automatic reorientation based on second degree surface fitting was in our hands less successful than reported in the literature. The tool using radial long-axis slices provides a better standard for testing reorientation algorithms than the traditional manual method. (orig.)