Transitioning New Technology into the Reading Room: A Reconstruction Approach for New Digital PET Users

1596 Objectives: Next-generation digital photon counting PET/CT systems have improved system sensitivity and time of flight timing resolution which enables more precise, higher definition imaging. The resulting improvements in image quality through high definition reconstruction protocols also leads to more accurate quantification. While this improvement is critical to advancing the utility of PET, there is a practical challenge of familiarizing oneself with the changes in both image quality and quantification compared to current experiences. As with other recent technology advances, such as the introduction of time of flight, a secondary reconstruction approach can be employed. We examined the use of a secondary reconstruction protocol for simulation of conventional image quality from a digitally acquired PET data set, via intra-individual comparison with images from a conventional imaging system. Materials and Methods: 65 patients underwent PET/CT imaging on a conventional system (Philips Gemini TF 64, cPET) approximately 75 minutes post-injection of 13 mCi 18F-FDG. Immediately before or after the conventional acquisition, a second acquisition was performed on a next-generation digital photon counting PET/CT system (Philips Vereos, dPET). Data were acquired from the top of the skull to the mid-thighs using 90 sec/bed position. Listmode data were reconstructed with each systems’ default settings: cPET - 4mm isometric voxel, 3 iterations, 33 subsets; dPET - high definition (HD) reconstruction with 2mm isometric voxel, 3 iterations, 11 subsets, point spread function correction and a 4.1 mm Gaussian filter applied. dPET data were also reconstructed with an protocol which meets EANM EARL criteria (conventional-equivalent (ce-dPET) reconstruction) - 4mm isometric voxel, 3 iterations 13 subsets, 5mm Gaussian filter, which was previously validated via phantom data. Regions of interest (ROIs) were placed over target lesions and in a variety of background tissues for quantitative comparison. Results: Visually, the cPET and ce-dPET images appeared very similar in quality and lesion detectability. Quantitatively the two reconstruction sets were well matched as well. The average SUVmax of target lesions for the default dPET reconstructions was 9.7. The cPET average SUVmax was only 6.5, as anticipated due to partial volume effects in smaller lesions. The ce-dPET reconstruction results were comparable to cPET with an average SUVmax of 6.6. The decrease was due to an increase in partial volume effects introduced by the use of a smaller reconstruction matrix in addition to the smoothing of the Gaussian filter alone. In background tissues, the SUVmean varied by less than 5% among all reconstruction settings. Conclusion: In this intra-individual comparison study we validated and presented the ability to produce conventional-equivalent digital PET images in order to provide quantitative and visual results consistent with conventional PET imaging. As new technology is introduced into clinical practice, a dual reconstruction approach can help reading physicians to familiarize themselves with changes inherent in the images, as well as to better assess imaging performed on a new digital system to older conventional photomultiplier tube system acquired images.