TY - JOUR T1 - Effects of Bayesian Penalized Likelihood Reconstruction on [11C]-Pittsburgh compound B time-of-flight PET/CT: A Preliminary Study JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 3022 LP - 3022 VL - 62 IS - supplement 1 AU - Shulin Yao AU - Baixuan Xu AU - Runze Wu Y1 - 2021/05/01 UR - http://jnm.snmjournals.org/content/62/supplement_1/3022.abstract N2 - 3022Introduction: [11C]-Pittsburgh compound B (PIB) PET is used for imaging amyloid plaques in patients with Alzheimer's disease (AD). PET acquisition is usually started at 40-60 minutes post-injection because PIB distribution reaches a stable state at that time window which is at 2-3 times of the half-life for 11C. Therefore, [11C]-PIB PET frequently works in a low-count condition and hence results in sub-optimal image quality. Bayesian penalized likelihood (BPL) reconstruction includes a penalization term in the iterations, which allows more accurate SUV quantification as well as lower noise than the ordered-subset expectation maximization (OSEM) reconstruction. Therefore, the aim of this study was to investigate the effects of BPL reconstruction on the visual image quality and semi-quantitative evaluation of [11C]-PIB PET/CT. Five subjects undergoing [11C]-PIB PET/CT were enrolled retrospectively, including 2 amyloid negative and 3 amyloid positive subjects. The patients received a quick injection of 310-421 mBq [11C]-PIB. After 38-43 min of waiting, list-mode brain PET data were acquired for 15 min on a PET/CT with 236 mm axial field-of-view of LYSO crystal coupling with photomultiplier tube (uMI510, United Imaging, China). PET images were reconstructed with our routine protocol (OSEM with 4 iterations and 24 subsets) using 15-min data. The list-mode data were split into 5 minutes and reconstructed with the BPL algorithm (R-OSEM, United Imaging, China) using a penalization of 0.2. Both OSEM and BPL reconstructions used 2.344x2.344x2.44 mm3 voxel size, time-of-flight (TOF), point spread function (PSF) model, and all necessary corrections. The visual image quality (VIQ) was scored by an experienced nuclear radiologist using a 5-point scale (1-unacceptable, 5-excellent). The reader reviewed the OSEM and BPL images in two sessions one month apart, respectively. The reading order was randomized to reduce memory effects. The semi-quantitative evaluation was performed by another nuclear radiologist using NeuroQ (Syntermed Inc., Atlanta, USA) that can obtain the cortex-to-whole cerebellum SUV ratios (SUVr) of 46 brain regions. The VIQ scores and SUVr of OSEM and BPL images were compared with Wilcoxon signed-rank test for the non-normality. The linear regression was performed to analyze the relationship of SUVr between OSEM and BPL reconstructions. The mean of VIQ scores was 2.6±0.5 and 3.6±0.5 for OSEM and BPL images (p < 0.05). The mean difference of SUVr was 0.006±0.01 between OSEM and BPL images (p = 0.43). The linear regression result was: SUVr_BPL = 1.04 × SUVr_OSEM - 0.076 with both p < 0.01 and adjusted R2 = 0.96. The BPL reconstruction improved the visual image quality with short acquisition time compared to OSEM reconstruction. The results of cortex-to-whole cerebellum SUVr were linearly correlated between OSEM and BPL images without statistically significant change. The BPL reconstruction can provide higher quality [11C]-PIB PET images with short scan time, which has the potential to increase patient comfort and reduce motion artifact. ER -