Numerical and human observer comparisons of tumor detection for 2D versus 3D whole-body PET imaging protocols

In this work we compared the impact of the acquisition mode (2D vs. 3D) on the performance of numerical and human observers in detecting and localizing tumors in whole-body PET imaging. We used a non-Monte Carlo simulator to generate multiple realizations of whole-body data with added 1cm-diameter spherical lesions (targets) and with accurate noise properties. The simulations were based on a Siemens/CTI HR+ scanner. The count rates were set to those corresponding to a 444 MBq (12 mCi) injected dose in 3D and 2D, which was shown to correspond to the peak NEC value in 3D for an average patient size, and we also analyzed the case equivalent to a 740 MBq (20 mCi) injected dose in 2D (2Dmax), which was considered as the maximum allowable dose based on considerations of dosimetry and cost. The simulated 2D data sets were reconstructed using 2D attenuation-weighted (AW) OSEM and the 3D data sets were reconstructed with FORE+AWOSEM