Simulating respiratory motion in whole-body PET imaging with the MCAT phantom

Respiratory motion is known to affect the quality of PET imaging of the thorax and abdomen, hindering the localization, detection of the lesions, and quantification of tracer uptake in a tumor. It also poses a challenge for the registration of PET and CT scans, even with a combined PET/CT scanner. Our long term goal is to study the impact of respiratory motion on lesion detection for FDG-PET scan. In this paper, we used the model proposed in the new dynamic MCAT phantom to simulate FDG-PET scans. In order to prepare the input data for the simulation, two sets of 16 temporal frames of the MCAT phantom were generated along a complete respiratory cycle. We generated an emission data set, using typical values of FDG uptake for each tissue, and a set of corresponding attenuation maps for 511 keV photons. After a thorough inspection of the different moving structures, we introduced 3 hot spherical lesions (7,10 and 12 mm diameter) in the liver, mediastinum and right lung, respectively with contrasts of 4:1, 8:1 and 13:1. We assumed that the liver and mediastinum approximately move as rigid bodies in the current phantom model. Both emission and attenuation data sets were input to an analytical PET simulator. The resulting analytical emission and transmission sinograms corresponding to each frame were averaged. Poisson noise was introduced in the emission projections and images were reconstructed using the FORE-AWOSEM algorithm, after correction for normalization and attenuation. The visual inspection of the images shows the lesions blurred and elongated in comparison with the images originating from a ´static´ frame. We also found that respiratory motion may preclude the detection of the smallest lesion inserted in the liver.