Performance evaluation of the Bayesian Penalized Likelihood Reconstruction Algorithm Q.Clear on BGO PET/CT system, according to NEMA NU2-2012 standard

2627 Objectives The purpose of this study was to evaluate the performance measurement value of PET / CT system Discovery-IQ (D-IQ) for 5-ring (GE Healthcare, Waukesha, WI) based on NEMA NU2-2012 procedures when images were reconstructed with a new algorithm called Q.Clear. Methods This D-IQ scanner, with 6.3(tangential) × 6.3(axial) × 30(radial) mm BGO crystal, includes a dual energy acquisition channel technology that reduces dead time losses and pileup at high count rates and should provide a high sensitivity response. The Q.Clear provided with D-IQ is the newest reconstruction algorithm, in which the image quality is determined by factor β of noise optimization. Performance measurements of the PET scanner were made using the NEMA NU2-2012 procedures for spatial resolution, scatter fraction, sensitivity, counts rate loss and random coincidence estimation, noise equivalent count rate (NECR) and image quality. As per NU2 2012 spatial resolution was calculated 1 cm center, at 10 cm and 20 cm off center in radial, tangential and axial directions. Sensitivity was calculated at center and 10 cm off center and system sensitivity was calculated by using these two values. Scatter fraction and NECR were measured. Also, image quality was evaluated with various β factors. Results The radial, tangential and axial FWHM were 4.65 mm, 4.52 mm and 4.85 mm at 1 cm off center, 5.85 mm, 5.16 mm and 4.96 mm at 10 cm off center, and 8.27 mm, 5.66 mm and 4.92 mm at 20 cm off center, respectively. The system sensitivity was 26.3 cps/kBq. The peak NECR was 113.3kcps at 9.5kBq/ml. Scatter fraction was 37.85%. Correction accuracy for the dead time losses and random event counts had a maximum absolute error of 3.55% below the NECR peak. In NEMA image quality, the hot contrast recovery values for 10, 13, 17, and 22 mm spheres were 42.4%, 54.5%, 67.1%, and 69.8%, respectively and cold contrast recovery values for 28 and 37 mm spheres were 72.5% and 74.4%, respectively; the corresponding background variability values for these spheres were 8.5%, 6.9%, 5.5%, 4.2%, 3.4%, and 2.6%, respectively. Recovery coefficients improved in a range of 7.1 to 11% depending on the lesions’ size (average of 9.2%) when images we reconstructed using Q.Clear (β=50). The contrast recovery values and background variability values decreased with the increase of factor β. Conclusions The D-IQ demonstrated high sensitivity and excellent physical performance. The new reconstruction algorithm Q.Clear allows us to get improved image quality; however, further investigations are required to determine the optimal factor β.