Evaluation of myocardial defect detection between parallel-hole and fan-beam SPECT using the Hotelling trace

The objective of this study was to implement the Hotelling trace (HT) to evaluate the potential increase in defect detection in myocardial SPECT using high-resolution fan-beam (HRF) versus parallel-hole (HRP) collimation and compare results to a previously reported human observer study [1]. Projection data from the 3D MCAT torso phantom were simulated including the effects of attenuation, collimator-detector response blurring and scatter. Poisson noise fluctuations were then simulated. The HRP and HRF collimators had the same spatial resolution at 20 cm. The total counts in the projection data sets were proportional to the detection efficiencies of the collimators and on the order of that found in clinical Tc-99m studies. In six left-ventricular defect locations, the HT found for HRF was superior to that for HRP collimation. For HRF collimation, the HT was calculated for reconstructed images using 64x64, 128x128 and 192x192 grid sizes. The results demonstrate substantial improvement in myocardial defect detection when the grid size was increased from 64x64 to 128x128 and slight improvement from 128x128 to 192x192. Also, the performance of the Hotelling observer in terms of the HT at the different grid sizes correlates at better than 0.95 to that found in human observers in a previously reported observer experiment and ROC study.