A channelized Hotelling observer study comparing cone-beam, fan-beam and parallel-hole collimation in /sup 99m/Tc-sestamibi myocardial SPECT with a heterogeneous phantom population

This study compared parallel-hole (PH), fan-beam (FB) and cone-beam (CB) collimation acquisition methods for myocardial perfusion, 99mTc-sestamibi myocardial SPECT imaging. For the task of myocardial defect detection, the channelized Hotelling observer (CHO) was used to compute the area under the ROC curve (AUC) in simulated SPECT images acquired using one of the following 5 single-detector acquisition methods: parallel-hole with 360deg acquisition (PH360), parallel-hole with 180deg acquisition (PH180), fan-beam with 360deg acquisition (FB360), fan-beam with 225deg acquisition (FB225), or cone-beam with 360deg acquisition (CB360). Unlike previous studies comparing PH, FB and CB, this study simulated a population of phantoms with variations in anatomy, defects and 99mTc-sestamibi uptake, as seen in actual cardiac SPECT patient images. Monte Carlo simulations were used to generate low noise projection data. Data were scaled to the same total scan time and to clinical count levels before adding Poisson noise. They were reconstructed using iterative OSEM with attenuation correction at various parameter settings. A single short axis (SA) slice was extracted, low-pass filtered at various cutoff frequencies and then rescaled to integer pixel values ranging from 0-255. For each acquisition method, OSEM parameter setting and filter cutoff, the AUC was estimated from 720 images. At the optimum parameter settings, the CB360 method yielded the highest AUC, followed by FB360, FB225, PH180 and PH360. The difference in AUC between the CB360 method and the other methods was statistically significant (p<0.05)