Impact of Bayesian penalized likelihood algorithm on the metabolism parameters and image quality in 18F-FDG PET/CT background tissues
10.3760/cma.j.cn321828-20200218-00050
- VernacularTitle:BPL重建技术对 18F-FDG PET/CT正常组织的代谢参数及图像质量的影响
- Author:
Lejun LIN
;
Yongtao WANG
;
Xiaoqin YAO
;
Xiaoning LYU
;
Jinyu LI
;
Shanchun LI
- From:
Chinese Journal of Nuclear Medicine and Molecular Imaging
2021;41(3):155-160
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the impact of Bayesian penalized likelihood (BPL) PET reconstruction method on the uptake of 18F-fluorodeoxyglucose (FDG) and signal to noise ratio (SNR) of lungs, liver, aorta and bones. Methods:From March 2019 to June 2019, the 18F-FDG PET/CT images of 60 patients with clinical diagnosed tumors (29 males, 31 females, age: 24-89 (60.4±15.2) years) in Yuhuangding Hospital were retrospectively analyzed. PET images were reconstructed with ordered subset expectation maximization (OSEM), time of flight (TOF)+ point spread function (PSF) and BPL (β=350) algorithms. Volumes of interest (VOIs) were delineated on the right upper lung lobe, the right liver, aortic root and lumbar vertebra. The mean standardized uptake value (SUV mean), maximum standardized uptake value (SUV max), peak of lean body standardized uptake value (SUL peak), standard deviation of standardized uptake value (SUV SD) and the SNR were measured. The percentage of SNR change (%ΔSNR) between the BPL method and non-BPL methods were calculated. The correlations between body mass index (BMI) and %ΔSNR were analyzed by Pearson correlation analysis. One-way analysis of variance and the least significant difference (LSD) t test were used to analyze the data. Results:There were no significant differences of SUV mean and SUL peak in lung, aorta, liver and lumbar vertebra among 3 methods ( F values: 0.04-1.95, all P>0.05). The SUV max in lung, aorta, liver and lumbar vertebra of BPL reconstruction (1.14±0.82, 2.13±0.37, 2.95±0.50 and 2.76±0.87) was significantly lower than those of TOF+ PSF (1.56±0.61, 2.99±0.75, 4.32±0.94 and 4.05±1.48) and OSEM (1.51±0.67, 3.00±0.70, 4.45±1.12 and 3.81±1.06) reconstructions ( F values: 20.59-52.24, all P<0.001) and SUV SD (0.13±0.07, 0.20±0.05, 0.26±0.06, 0.38±0.17) was also significantly lower than those of TOF+ PSF (0.24±0.11, 0.43±0.11, 0.58±0.15, 0.67±0.21) and OSEM (0.21±0.09, 0.42±0.10, 0.58±0.14, 0.63±0.20) reconstructions ( F values: 24.46-124.95, all P<0.001), while the SNR (4.67±1.34, 7.74±2.22, 8.17±1.77, 4.45±1.22) was significantly higher than those of TOF+ PSF (2.54±0.72, 3.55±0.82, 3.77±0.91, 2.49±0.69) and OSEM (2.65±0.64, 3.67±0.80, 3.75±0.87, 2.60±0.67) reconstructions ( F values: 83.04-247.73, all P<0.001). However, there were no significant differences between OSEM and TOF+ PSF reconstructed images in SUL peak, SUV mean, SUV SD and SNR (all P>0.05). In BPL group, SNR increased with the increase of BMI, and there were statistically differences of aortic SNR (7.07±2.21 vs 9.67±2.26) and liver SNR (7.75±1.85 vs 9.32±0.70) between BMI<25 kg/m 2 and BMI≥30 kg/m 2 ( F values: 3.46 and 4.19, both P<0.05). Positive correlations were found between %ΔSNR of lung, aorta, liver and lumbar vertebra in OSEM and TOF+ PSF and BMI ( r value: 0.042-0.354, all P<0.05). Conclusion:In background tissues, BPL algorithm has no significant impact on absolute quantification compared with OSEM and TOF + PSF reconstruction methods but it can significantly improve SNR, especially for the patients with large body weight.
