Quantitative and Qualitative Performance Evaluation of Maximum a posteriori PET reconstruction algorithm (Q.Clear) using Body and Brain PET Phantom Images

1811 Objectives Conventionally, there is trade-off between SNR (signal-to-noise ratio) and SUV accuracy; better SNR lead to underestimated SUVmax; better SUVmax recovery lead to lower SNR. Q.Clear (maximum a posteriori (MAP) PET reconstruction) can enhance this trade-off condition using a priori information of neighbor-voxels. To demonstrate this, we examined Hoffman 3D Brain Phantom (HP) and NEMA IEC Phantom (NP) images reconstructed by both reconstruction methods qualitatively and quantitatively. Methods HP/NP PET images were acquired on Disovery 690 for simulating clinical brain/body PET. Conventional 3D OSEM with/without Time-of-Flight (TOF) information and/or point spread function (PSF) information were also used for the comparison. For quantitative comparison we used recovery coefficient (RC) and SNR of NP data (varying MAP neighborhood smoothing paramter as 50/150/250). For qualitative comparison two board-certified physicians (MO and SJ) assessed image quality of HP with respect to Spatial Resolution (ability to detect fine structure), Tissue Contrasts (ability to differentiate parenchyma/background and gray/white matter), and SNR. Results The RCs by MAP with/without TOF were 5/5.2, 4/4.5, and 3.7/4.2, respectively. SNRs were 9/8, 17/16, and 23/21, respectively. Taken together, both RC/SNR were similar between MAP with/without TOF (unlike the SNR of conventional method with/without TOF: 3/16). At similar RCs, SNR was 5-fold increased by MAP. Expert visual inspection of HP accorded with this; conventional method resulted in heterogeneous basal ganglia and uncertain edge. All image quality elements were good for MAP algorithm with low kernel parameter. Conclusions MAP reconstruction outperformed than conventional reconstruction algorithms in terms of SNR at similar SUV recovery, according with expert visual inspection. Therefore, Q.Clear reconstruction may advance image quality of clinical PET.