Quantitative evaluation of reconstruction algorithms for motion compensated PET

The objective of this work was to investigate the effects different reconstruction algorithms have on lesion detection and quantification in respiratory gated and motion corrected positron emission tomography (PET) images. The NCAT phantom was simulated with respiratory motion, and 10 mm lesions with a source to background ratio of 5 to 1 or 10 to 1 were placed at the apex, middle, and base of the right lung. The NCAT data were binned into 16 gates, filtered with a smoothing kernel to represent PET system resolution, and then forward projected to create sinograms. Data were reconstructed with four different algorithms: FBP, OSEM 2 iterations and 8 subsets, OSEM 4 iterations and 8 subsets, and OSEM 8 iterations and 8 subsets. In addition to gating, a motion correction method based on rigid image registration of the gated images was used on the data. The mean lesion signal, lesion noise, and lesion contrast to noise ratio (CNR) were evaluated.