《中国医学装备》杂志社

目的：分析块序正则化最大期望值(BSREM)重建算法不同β值重建对¹⁸F-氟化铝-前列腺特异性膜抗原-11(Al¹⁸FPSMA-11)PET/CT图像质量参数和半定量参数的影响，经Al¹⁸F-PSMA-11显像剂获取最佳图像质量和准确半定量参数的β值。方法:回顾性分析2022年7月至2023年11月就诊于天津医科大学肿瘤医院，经病理诊断为前列腺癌并且在本院行Al¹⁸F-PSMA-11 PET/CT检查的23例患者的56个病灶，所有图像分别选择有序子集最大化(OSEM)算法重建和BSREM算法重建。BSREM选择不同的β值：β100、β200、β300、β400、β500、β600、β700、β800、β900和β1 000进行重建。视觉分析由两名医师对所有图像进行独立评分。测量两种重建方法后的图像噪声(Noise)、信噪比(SNR)、信号背景比(SBR)、最大标准化摄取值(SUV_max)。分析β100～β1000的Noise、SNR、SBR及SUV_max变化，并结合视觉分析结果评价Al¹⁸F-PSMA-11 PET/CT BSREM重建β值对图像质量和半定量参数的影响。结果:23例患者BSREM重建算法图像随β值上升Noise水平整体呈现下降趋势(Q=227.13,P<0.001),β100和β200时Noise水平高于OSEM重建(Z=-4.20,-2.54,P<0.05)。随β值上升所有病灶的SNR整体呈递增趋势(Q=305.33,P<0.001),β100时SNR与OSEM重建的SNR比较差异无统计学意义(P>0.05),β200～β1 000时SNR高于OSEM重建，差异均有统计学意义(Z=-2.446,-3.920,-5.839,-6.089,-6.464,-6.446,-6.196,-5.562,-5.821,P<0.05)。β600～β1 000相邻值间SNR比较差异均无统计学意义(P>0.05),SNR趋于平稳。所有病灶SBR随β值上升逐渐下降，β600时与OSEM重建SBR差异无统计学意义(P>0.05)。β700～β1 000时SBR低于OSEM重建，差异均有统计学意义(Z=-1.098,-2.196,-3.250,-4.321, P<0.05),β700～β1 000相邻值间SBR比较差异均无统计学意义(P>0.05),SBR趋于平稳。SUV_max随β值增加逐渐下降(Q=529.99,P<0.001),β100～β500时SUV_max高于OSEM重建，差异均有统计学意义(Z=-5.607,-4.714,-3.589,-2.625,-1.357,P<0.05)。β600和β700时SUV_max与OSEM重建比较差异无统计学意义(P>0.05)。β800～β1 000时SUV_max低于OSEM重建(Z=-2.107,-3.188,-4.268,P<0.05)。视觉评分结果显示，BSREM重建算法β100～β300、β900和β1 000图像质量低于OSEM重建，β400和β800图像质量近似于OSEM重建，β500～β700图像质量显著优于OSEM重建。结论：BSREM重建β值影响Al¹⁸F-PSMA-11 PET/CT图像Noise、SNR、SBR及SUV_max，且影响图像质量和诊断水平，而最适重建β值建议为600。

Objective: To investigate the effect of different β values of the Block Sequential Regularized Expectation Maximization(BSREM) algorithm on the image quality parameters and semi-quantitative parameters of [¹⁸F]aluminum fluoride-prostate-specific membrane antigen-11(Al¹⁸F-PSMA-11) PET/CT, and to screen out the optimal reconstruction βvalue for this imaging agent. Methods: A retrospective study was conducted on 23 patients pathologically diagnosed with prostate cancer who underwent Al¹⁸FPSMA-11 PET/CT examination at Tianjin Medical University Cancer Institute and Hospital from July 2022 to November 2023, with a total of 56 lesions enrolled. All images were reconstructed using the Ordered Subsets Expectation Maximization(OSEM) algorithm and the BSREM algorithm respectively. For the BSREM algorithm, the β values were set as a gradient series of 100, 200, 300, 400, 500, 600, 700, 800, 900 and 1000. Two senior radiologists performed double-blind visual quality scoring of all images. The image noise(Noise), signal-to-noise ratio(SNR), signal-to-background ratio(SBR) and maximum standardized uptake value(SUV_max) were measured after reconstruction with the two algorithms. The variations of Noise, SNR, SBR and SUV_max with β values ranging from 100 to 1000 were analyzed, and combined with the results of visual analysis, the effect of β values in BSREM reconstruction on the image quality and semi-quantitative parameters of Al¹⁸F-PSMA-11 PET/CT was evaluated. Results: In 23 patients, the overall noise level decreased with increasing β values(Q=227.13, P<0.001), while the noise levels at β100 and β200 were higher than those reconstructed by OSEM(Z=-4.20,-2.54, P<0.05). The signal-to-noise ratio(SNR) of all lesions showed an overall increasing trend with rising β values(Q=305.33, P<0.05). At β100, the SNR did not differ statistically from OSEM reconstruction(P>0.05). Between β200 and β1000, the SNR was higher than that reconstructed by OSEM(Z=-2.446,-3.920,-5.839,-6.089,-6.464,-6.446,-6.196,-5.562,-5.821, P<0.05). Between adjacent values at β600 and β1000, the SNR did not differ statistically(P>0.05), indicating a plateau in SNR.The SBR of all lesions gradually decreased with increasing β values. At β600, the SBR showed no difference from OSEM-reconstructed SBR(P>0.05). Between β700 and β1000, the SBR was lower than OSEM-reconstructed SBR(Z=-1.098,-2.196,-3.250,-4.321, P<0.05). There was no statistically significant difference in SBR between adjacent values within β700 to β1 000(P>0.05), and SBR tended to stabilize. The SUV_max gradually decreased with increasing β values(Q=529.99, P<0.001). Between β100 and β500, the SUV_max was higher than OSEM-reconstructed(Z=-5.607,-4.714,-3.589,-2.625,-1.357, P<0.05). At β600 and β700, the SUV_max showed no statistically significant difference compared to OSEM-reconstructed(P>0.05). Between β800 and β1 000, the SUV_max was lower than OSEM-reconstructed(Z=-2.107,-3.188,-4.268, P<0.05). Visual assessment results indicated that image quality at β100 to β300, β900, and β1000 was inferior to OSEM-reconstructed, while β400 and β800 were comparable to OSEM-reconstructed. Image quality at β500 to β700 was significantly superior to OSEM-reconstructed. Conclusion: The β value of BSREM reconstruction has significant impact on the Noise, SNR, SBR and SUV_max of Al¹⁸F-PSMA-11 PET/CT images, which can affect the quality of image and the diagnostic level. It is recommended that β=600 is selected as the optimal reconstruction β value.