Phantom Evaluation of a Post Reconstruction Misregistration Correction Technique for Cardiac PET

374 Introduction: Misregistration artifacts in cardiovascular PET imaging are one of the greatest sources on misdiagnosis. However, most misregistration correction (MRC) algorithms in use rely on shifting the transmission and emission sinogram data prior to reconstruction. Correction for misregistration typically must be performed prior to reconstruction. Because many systems either cannot perform misregistration correction or cannot archive raw sinogram data, there is a need to perform misregistration correction on the post-reconstruction tomographic data. We examined the effectiveness of a post-reconstruction (MRC) algorithm using phantom acquisitions with simulated misregistration. Methods: A Data Spectrum Anthropomorphic Phantom was prepared with F-18. The activity concentration and acquisition times were similar to a Rb-82 cardiac PET study. The studies were acquired on a Siemens Biograph 16 PET/CT system, using the system’s misregistration correction software to simulate varying amounts of MRC (-15, -10, -5, 5, 10, 15 mm, lateral and transverse directions). Post-reconstruction MRC was applied using a previously reported iterative reprojection algorithm (ImagenUniversal, CVIT Kansas City). This algorithm combines reconstructed emission data, transmission data and MRC. 17-segment maps were constructed from the reconstructed tomograms and analyzed on a semi-quantitative scale (0-100% of maximum pixel). Pre- and post-reconstruction MRC images were compared to a reconstruction without misregistration. Results: Uncorrected misregistration artifacts reduced lateral wall count by up to 21% for lateral shifts and 15%: Misregistration artifacts in lateral directions were corrected using both pre- and post-MRC (10%, 9% respectively). Misregistration artifacts resulting from transverse shifts had similar results, 9% for both pre- and post-MRC. Mean changes in myocardial counts are in Figure 1: Segment by segment analysis of differences of pre- and post-reconstruction MRC demonstrated only three statistically significant shifts (p<0.05): two showing improvement for pre-MRC (-5,0) and (0,10) and one improvement for post-MRC (0,-15). Conclusions: The post-reconstruction MRC examined produced reconstructions that are comparable to pre-reconstruction MRC images. These results were consistent for lateral and transverse shifts up to 15 mm. The availability of post reconstruction misregistration correction can allow for greater flexibility in applying this important correction without the need for transferring or saving the raw sinogram data.