Joel Fredrik Astrup Aanerud

Validation of PSF-based 3D reconstruction for myocardial blood flow measurements with Rb-82 PET

Research output: Contribution to conferencePosterResearchpeer-review

Aim:The use of PSF-based 3D reconstruction algorithms (PSF) is desirable in most clinical PET-exams due to their superior image quality. Rb-82 cardiac PET is inherently noisy due to short half-life and prompt gammas and would presumably benefit from PSF. However, the quantitative behavior of PSF is not well validated and problems with both edge-effects and unphysical contrast-recovery have been reported.1 In this study, we compare myocardial blood flow (MBF) and coronary flow reserve (CFR) obtained using GEs implementation of PSF, SharpIR, with the conventional method for reconstruction of dynamic images, filtered backprojection (FBP). Furthermore, since myocardial segmentation might be affected by image quality, two different approaches to segmentation implemented in standard software (Carimas (Turku PET Centre) and QPET (Cedar Sinai)) are utilized. Method:14 dynamic rest-stress Rb-82 patient-scans performed on a GE Discovery 690 PET/CT were included. Images were reconstructed in an isotropic matrix (3.27x3.27x3.27 mm) using PSF (SharpIR: 3 iterations and 21 subsets) and FBP (FORE FBP) with the same edge-preserving filter (3D Butterworth: cut-off 10 mm, power 10). Analysis: The dynamic PET-data was analyzed using a 1-tissue model corrected for incomplete extraction.2 Two different approaches to orientation and segmentation were performed: 1) individual segmentation (IN) of PSF and FBP data (in QPET), 2) identical segmentation (ID) of PSF and FBP data using the same ROIs on both datasets (in Carimas). Results: Good correlation between values of MBF was found regardless of segmentation approach for both global (R2=0.98 in both cases) and segmental data (ID: R2=0.95, IN: R2=0.94). This was also the case for global CFR (ID: R2=0.96, IN: R2=0.89) and segmental CFR (ID: R2=0.89, IN: R2=0.85). However, PSF results were slightly higher than FBP when using identical segmentation (Global: 10±6%, p>0.05, Segment: 8±2%, p>0.05). This was not observed when using individual segmentation. Conclusion: Good correlation between PSF and FBP MBF and CFR values on both global and segmental level was found. The slightly higher MBF for PSF than FBP when using identical segments can be understood in terms of better resolution using PSF (less spill-over between myocardium and input ROIs). The difference of less than 10% must, however, is most likely clinically irrelevant. Thus, the data suggest that SharpIR reconstruction can be used for myocardial blood flow measurements with Rb-82. References:1) Rapisarda et al, Phys Med Biol 2010;55:4131-4151 2) Lortie et al, Eur J Nucl Med Mol Imaging 2007;34:1765-1774
Original languageEnglish
Publication year2013
Publication statusPublished - 2013
EventEANM '13 - Lyon, France
Duration: 19 Oct 201323 Oct 2013


ConferenceEANM '13

See relations at Aarhus University Citationformats

ID: 76800127