Effect of Different Reconstruction Algorithms on the Dynamics and Modeling Parameters of 18F-Galacto-RGD in mice

摘要

So far FBP (Filtered Back Projection) is still the standard reconstruction method for kinetic modeling of PET data. Iterative reconstruction algorithms, such as OSEM (Ordered Subsets Expectation Maximization) or MAP (Maximum A Posteriori) create images with better signal-to-noise ratios and less streak artifacts. This is of special interest in the short-time frames with low statistics at the beginning of a dynamic acquisition. To determine if iteratively reconstructed PET data could be used for kinetic modeling of pre-clinical data, we compared OSEM2D and OSEM3D/MAP data with FBP data using microPET data. Nude mice (n=6) bearing a human melanoma tumor were imaged over 90 minutes with a FOCUS 120 microPET. We applied 18F-Galacto-RGD via a tail vein catheter into the animal (9-19 MBq). 18F-Galacto-RGD is used for imaging the α{sub}vβ{sub}3 expression during angiogenesis and on specific tumor cells. Each list mode dataset was binned into 19 frames and reconstructed with FBP, OSEM2D and OSEM3D/MAP. The data was not corrected for scatter and attenuation. Volumes of interest (VOIs) were drawn on the heart and tumor tissue. No corrections for spillover and partial volume effect were applied. For the kinetic modeling of the tumor data a standard 2-tissue-compartment model with reversible binding was used (fitting parameters: K{sub}1, k{sub}2, k{sub}3, k{sub}4 (rate constants) and D{sub}v (vascular blood fraction)). The input function TACs of the iteratively reconstructed data showed no considerable difference to the FBP data. Only the OSEM3D/MAP TACs for the tumor VOIs were clearly and consistently higher for all datasets than the corresponding FBP and OSEM2D TACs. Kinetic modeling shows negligible parameter changes for K{sub}1, k{sub}2, k{sub}3 and k{sub}4 for all reconstruction algorithms. The higher OSEM3D/MAP tumor values are compensated only with an overestimated D{sub}v (D{sub}v(FBP)=2.2%, D{sub}v(OSEM3D/MAP)=3.3%). Further work needs to be done on optimizing the iteration settings for the OSEM3D/MAP.