High definition digital photon counting PET for neurology: A phantom-based optimization approach

Objectives: The next-generation of PET/CT systems with digital photon counting allow for consideration of higher definition reconstruction, in addition to lower dose, faster acquisition imaging, than has been the standard to now. These advantages are of particular interest for neurologic imaging where improved image quality would benefit current and future applications of PET. The aim of this study was to evaluate the application of high definition image reconstruction methodologies provided by digital photon counting PET (dPET) in neurologic imaging data. Materials and Methods: A Hoffman brain phantom was filled with 0.8 ¹⁸F in the background volume and four spheres with volumes ranging from 2ml to .125 ml were filled with ¹⁸F to a range of 3-4:1 hot:background ratios, as well as two spheres filled with distilled water. The spheres were placed throughout the volume of the phantom. The phantom was imaged on a digital photon counting PET/CT (Philips Vereos). Ten runs of five minute acquisitions were repeated with the brain 256 mm field of view (FOV) and the whole body 576 mm FOV used alternately. Each acquisition was reconstructed using listmode data with high definition (HD) 2x2x2 mm voxel volumes / 288x288 matrix with 3 iterations and subsets ranging from: 21, 17, 13, and 9. Point spread function (PSF) only, Gaussian only, a combination of both or no filters were enabled. Sixteen regions of interests (ROIs) were placed in background regions as well as over the each sphere.

Results: After decay correction, the results from the five runs performed with each FOV were very consistent for any given reconstruction parameter choice. Averaging the measurements for all runs with each FOV then revealed very similar results between either the 256 or the 576 FOV for ROIs placed in the background/gray matter regions. In the hot spheres however, the quantitative results were varied dependent upon the sphere size. The smaller spheres had activity concentrations measured more accurately with the brain 256 FOV than the whole body 576 FOV. This disparity could be overcome in part by use of the PSF correction alone in the 576 FOV reconstructions. For the brain 256 FOV, the average recovery coefficients (RCs) of the hot spheres were similar within reconstructions with different subsets, averaging 1.19, 1.20, 1.20, and 1.21 for reconstructions with 9, 13, 17, and 21 subsets each, respectively. For the whole body 576 FOV the RCs were again similar regardless of subset choice, averaging 0.93, 0.95, 0.95, and 0.95 for 9, 13, 17, and 21 subsets, respectively. Conclusion: Next-generation digital photon counting PET/CT allows for more precise imaging of the brain with accurate and robust quantification. Phantom analysis showed that excellent image quality can be achieved using both brain and whole body field of view by adjusting reconstruction settings, such as the number of subsets per iteration. Quantitative accuracy is maintained over the range of reconstruction settings, due in part to the quality of counts in the PET data resulting from digital photon counting. Research Support: ODSA TECH 13-060