Comparison of Myocardial Blood Flow quantification for 13NH3 and 3D-PET by using different image reconstruction algorithms (3D-Reprojection, 3D-OSEM, 3D-OSEM with Time of-Fligth and 3D-Point-Spread-Function) and De-Grado Model.

Aim: To compare the absolute quantification of Myocardial Blood Flow (MBF) with 13NH3 and 3D-PET by using different image reconstruction algorithms (3D-Reprojection, 3D-OSEM, 3D-OSEM with Time-of-Fligth and 3D-PET Point-Spread-Function) and De-Grado Model. Materials and methods: 22 patients with different cardiomyopathies (hypertensive, ischemic, hypertrophic and dilatative) underwent a 3D-PET (Discovery-690 ; GEMS, Milwaukee, USA) at rest and during vasodilatory stress after dipyradamole infusion (0.56 mg/kg/4 min). After the injection of 370 MBq of 13NH3, a 3D-PET dynamic scan was acquired with framing: 9x10s, 6x15s, 3x20s, 2x30s, and 1x900s. Off-line the 3D-PET data were reconstructed by using different reconstruction algorithms and parameter configurations as follow: a)3D-Reprojection, with the smallest possible filter imposed by the Nyquist limit, 3D-OSEM with two different number of iterations: b) 3 iterations and c) 5 iterations, and d)3D-OSEM accounting also for Time-Of-Flight and 3D-PET Point-Spread-Function. Quantification was performed with PMOD software (v3.1, PMOD Technologies, Zurich) using the De-Grado 1-compartment model. Two analysis were conducted. At first, VOIs of the blood pool and myocardial tissue were drawn on images a; subsequently they were applied onto reconstructions b, c, and d. In the second analysis, new VOIs were drawn on each differently reconstructed image set, in order to evaluate whether the reconstruction method could affect the the delineation of the VOIs. Comparison were made by Bland Altman plots. Results: The analysis performed using the same VOIs showed no significant mean difference between method a (used as reference) b and c. With method d a mean difference of +5% was observed. Standard deviations between the chosen reconstruction and method a were [Rest,Stress]: b[4.5%,6.5%], c[4.2%,6.1%], d[5.5%,9.2%]. Using different VOIs the analysis of mean differences gave identical results. The standard deviations, indicating the inter-algorithm variability, increased to b[8.2%,6.3%], c[7.1%,7.7%], d[7%,9.1%]. The spill-over fraction was minimized by method d: -20% compared to a and -31% compared to b. Method c had the same values as a. Conclusion: Absolute quantification of MBF with 13NH3, 3D-PET and De-Grado model is consistently stable across different image reconstruction methods. The most innovative reconstruction algorithm, including Time-Of-Flight information and 3D-PET Point-Spread-Function, is slightly biased compared with 3D-RP and needs less spillover correction. The between-algorithms variability remained limited to less than 9%. Thus, MBF quantification can be considered clinically robust to different image reconstruction techniques.