Skip to Main Content
The aim of the study was to determine the accuracy with which spillover and the partial volume effect (PVE) compensation can restore the local concentration of activity in the myocardium during cardiac perfusion imaging when co-registered high-spatial resolution anatomical information from a state-of-the-art CT system is available. We evaluated a voxel-specific template-projection-reconstruction algorithm which uses a 3D binary image of each organ segmented from aligned anatomical maps to generate projections that modeled the physical characteristics of the gamma camera excluding Compton scatter. Voxel specific compensation for spillover is obtained by subtracting from the voxel count in the heart the spillover fraction obtained from reconstruction of the neighboring structures times the average count in each neighboring structure. The counts in the heart are then divided by the voxel specific recovery coefficient for the heart obtained from the reconstruction of the projections of the heart anatomy, weighted by the fractional presence of an organ in the emission voxels determined by the anatomy. Investigation was made of MCAT phantoms with a normal LV distribution, and also with impaired blood flow in the inferior and anterior walls of the LV. The rescaled block iterative (RBI) reconstruction of data included compensation for attenuation, Compton scatter, and distance dependent resolution. Circumferential profiles were used to evaluate the accuracy of the method. We observed a significant improvement in the accuracy of the estimated concentration in reconstructed slices of normal myocardium. In areas of impaired blood flow (decreased concentration), the algorithm failed to remove the bias created by the spillover of activity from within the myocardial wall. A further bias is introduced when the areas of impaired flow also exhibit a decrease in motion and thickening in the CT slices. Spillover and partial volume compensation evaluated in this study is an excellent m- ethod to accurately represent an increase in activity concentration compared to background (hot spot imaging), but there is a residual bias when the activity concentration is lower than its immediate surroundings (cold spot imaging).