Monte Carlo Simulation Study of Several Camera Designs for the PET Component of a Dedicated Breast PET/CT Scanner

In this study we used Monte Carlo simulations to analyze the patient imaging performance of a range of camera geometries for the PET portion of a dedicated pendant breast PET/CT system. Performance was assessed by comparing NEC, rates, and scatter fractions for a range of injected FDG activities. The same LSO block detector (9times9 arrays of 3 mm times 3 mm times 20 mm crystals and 3.3 mm pitch) was used to compose a cylindrical, split-ring, five-sided box, and a planar dual-headed system. To simulate the patient an anthropomorphic model was created consisting of the NCAT phantom, volumes acting as the brain and bladder, and one of three CT-derived pendant breast shapes (representative of small, medium, and large volume breasts) appended to the left chest wall, with FDG concentration ratios set to those given by Ramos et al. Systems were modeled with LSO background, dead time effects (256 ns paralysable at the block level, and non-paralysable), a 7.5 ns coincidence time window, and energy window of 350 to 650 keV. The cylindrical scanner gave significantly higher NEC rates across all viable injection activities (~5 to ~20 mCi) regardless of the breast size imaged. For the medium sized breast, at an injection activity of 20 mCi, the cylindrical scanner showed a 3-fold gain in NEC rate compared to the dual-headed system (32.4 vs. 10.5 kcps) at the cost of a slightly higher scatter fraction (0.28 vs. 0.24). Additional simulations with an alternative cylindrical scanner system composed of DOI capable detectors (APD front end and PMT back) gave somewhat lower NEC rates compared with the conventional cylindrical design due to a lower effective packing fraction. Based on these results the cylindrical scanner was found the optimal geometry for prone dedicated breast PET/CT imaging.