The Impact of the Bayesian Regularization Likelihood Algorithm on the Semi-Quantitative Parameters of Pulmonary Nodules in <sup>18</sup>F-FDG PET/CT under Different Respiratory States

目的：探讨贝叶斯正则化似然(BPL)重建算法在不同呼吸状态下对18F-FDG PET/CT肺结节半定量参数的影响。方法：回顾性分析2022年6月至2022年10月间于青岛大学附属医院行18F-FDG PET/CT全身扫描的108例患者，共计139个肺结节。采用自由呼吸OSEM、自由呼吸BPL、屏气OSEM和屏气BPL重建算法对肺结节PET图像进行重建，比较组间最大标准摄取值(SUVmax)、平均标准摄取值(SUVmean)、峰值标准摄取值(SUVpeak)、信号/本底比值(SBR)等半定量参数。组间差异采用Wilcoxon秩和检验比较分析。结果：屏气OSEM组重建后摄取代谢参数SUVmax、SUVmean、SUVpeak和SBR高于自由呼吸OSEM组，分别是4.65 (2.38, 7.47)和3.02 (1.67, 5.61) (z = −9.53, p Objective: To investigate the impact of the Bayesian Penalized Likelihood (BPL) reconstruction algorithm on the semi-quantitative parameters of lung nodules in 18F-FDG PET/CT under different respiratory states. Methods: A retrospective analysis was conducted on 108 patients with a total of 139 lung nodules who underwent 18F-FDG PET/CT whole-body scans at the Affiliated Hospital of Qingdao University from June 2022 to October 2022. PET images of lung nodules were reconstructed using four algorithms: free-breathing OSEM, free-breathing BPL, breath-hold OSEM, and breath-hold BPL. The semi-quantitative parameters, including maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), peak standardized uptake value (SUVpeak), and signal-to-background ratio (SBR), were compared among the groups. Intergroup differences were analyzed using the Wilcoxon rank-sum test. Results: The SUVmax, SUVmean, SUVpeak, and SBR values in the breath-hold OSEM group were higher than those in the free-breathing OSEM group, with values of 4.65 (2.38, 7.47) vs. 3.02 (1.67, 5.61) (z = −9.53, p < 0.001), 2.71 (1.85, 3.56) vs. 2.25 (1.36, 3.12) (z = −8.67, p < 0.001), 2.65 (1.38, 5.52) vs. 2.05 (1.29, 4.01) (z = −7.75, p < 0.001), and 3.02 (1.45, 5.32) vs. 1.77 (1.01, 3.50) (z = −9.74, p < 0.001), respectively. The SUVmax, SUVmean, SUVpeak, and SBR values in the breath-hold BPL group were higher than those in the free-breathing BPL group, with values of 6.8 (3.51, 11.55) vs. 4.45 (2.19, 7.31) (z = −9.99, p < 0.001), 3.42 (2.52, 4.91) vs. 2.71 (1.68, 3.55) (z = −9.30, p < 0.001), 3.70 (2.07, 6.61) vs. 2.69 (1.72, 5.27) (z = −9.30, p < 0.001), and 4.44 (2.24, 7.61) vs. 2.59 (1.28, 4.66) (z = −10.07, p < 0.001), respectively. The SUVmax, SUVmean, SUVpeak, and SBR values in the BPL groups were higher than those in the OSEM groups under both breath-hold and free-breathing conditions (all p < 0.001). The breath-hold BPL group had a significantly greater impact on the metabolic parameters (SUVmax, SUVmean, SUVpeak, and SBR) and their changes (%ΔSUVmax, %ΔSUVmean, %ΔSUVpeak, and %ΔSBR) compared to the free-breathing BPL group (p < 0.001). Conclusion: The breath-hold BPL reconstruction algorithm reduces respiratory motion artifacts, significantly enhancing the detection rate of semi-quantitative parameters in PET images of pulmonary nodules, thereby providing a more precise imaging basis for the diagnosis of their benign or malignant nature.