Evaluation of 3D OSEM TOF vs PLEM algorithms for whole body 68GA-DOTATATE PET/CT studies.

1471 Objectives: The image quality and quantitative accuracy of PET depend on several factors such as uptake time, scanner characteristics and image reconstruction methods. For image reconstruction, the ordered subset expectation maximization (OSEM) statistical method is considered as gold standard. Recently, a new reconstruction method, the penalized-likelihood estimation (PLEM) algorithm, allowing for fully convergent iterative reconstruction with improved quantitation accuracy, has been introduced. Q.Clear (PLEM algorithm of GE Healthcare), uses as the only user-input variable a positive regularization parameter, named β. In the present study, we aim to compare quantitatively and visually the performance of Q.Clear for different values of β and different acquisition times per bed position against 3D OSEM TOF for 68Ga-DOTATATE examinations. Methods: Ten consecutive clinical whole-body 68Ga-DOTATATE examinations acquired on a digital Discovery MI TOF PET/CT (GE Healthcare) were included. The data were reconstructed using 3D OSEM TOF with clinical settings (1.5 min/FOV, 3 iterations, 8 subsets, 6 mm Gaussian filter) and Q.Clear with different β values and different acquisition time per bed position (1.5 min/FOV: β=300, 400, ⋯,1200; 1.0 min/FOV: β=600, 700, ⋯,1500 and 0.5 min/FOV: β=800, 900, ⋯,2500). Evaluation was performed on a phantom (Contrast Noise Ratio and Background Variability) and on clinical data (signal-to-noise ratio (SNR), signal-to-background ratio (SBR) and Noise). Finally, 2 experienced nuclear medicine physicians blinded to the variables have assessed visually contrast, sharpness, noise, liver homogeneity, tumor detectability and overall image quality. Results were reported on a scale of 1 (very poor) to 4 (very good). Results: On phantom evaluation optimal values of β were 900, 1200 and 2000 for 1.5 min/FOV, 1.0 min/FOV and 0.5 min/FOV acquisitions, respectively. Evaluation of clinical images reconstructed with the latter acquisition times using Q.Clear with β = 1100, 1300 and 2200 respectively, resulted in noise equivalent to 3D TOF OSEM with 1.5 min/FOV with an increase in SUVmax (14%, 9% and 4%, respectively), SNR (10%, 8% and 3%, respectively) and SBR (8%, 9% and 3%, respectively). Visual assessment resulted in the same β optimal values than the clinical images evaluation (β = 1100 , 1300 and 2200 for 1.5, 1.0 and 0.5 min/FOV, repectively) with a score for OSEM vs Q.Clear (1.5 min/FOV, 1.0 min/FOV and 0.5 min/FOV) of 3.3 vs 3.6, 3.4 vs 3.7 and 3.5 vs 3.7, respectively. Conclusions: Q.Clear reconstruction resulted in increased tumor SUVmax and in improved SNR and SBR at a similar level of noise compared to TOF OSEM, regardless of the acquisition time per bed position. The optimal value of β for 1.5 min/FOV was 1100. Q.Clear allowed for a shorter acquisition resulting in similar image quality to 3D OSEM TOF with 1.5 min/FOV when using β =1300 for 1.0 min/FOV and β =2200 for 0.5 min/FOV (see fig. 1).