Thuillier Philippe, Bourhis David, Pavoine Mathieu, Metges Jean-Philippe, Le Pennec Romain, Schick Ulrike, Blanc-Béguin Frédérique, Hennebicq Simon, Salaun Pierre-Yves, Kerlan Véronique, Karakatsanis Nicolas A, Abgral Ronan
Department of Endocrinology, University Hospital of Brest, Brest, France.
UMR 1304 Inserm GETBO, University Hospital of Brest, Brest, France.
Front Nucl Med. 2022 Sep 21;2:941848. doi: 10.3389/fnume.2022.941848. eCollection 2022.
To validate a population-based input function (PBIF) model that alleviates the need for scanning since injection time in dynamic whole-body (WBdyn) PET.
Thirty-seven patients with suspected/known well-differentiated neuroendocrine tumors were included (GAPETNET trial NTC03576040). All WBdyn 68Ga-DOTATOC-PET/CT acquisitions were performed on a digital PET system (one heart-centered 6 min-step followed by nine WB-passes). The PBIF model was built from 20 image-derived input functions (IDIFs) obtained from a respective number of patients' WBdyn exams using an automated left-ventricle segmentation tool. All IDIF peaks were aligned to the median time-to-peak, normalized to patient weight and administrated activity, and then fitted to an exponential model function. PBIF was then applied to 17 independent patient studies by scaling it to match the respective IDIF section at 20-55 min post-injection time windows corresponding to WB-passes 3-7. The ratio of area under the curves (AUCs) of IDIFs and PBIF were compared using a Bland-Altman analysis (mean bias ± SD). The Patlak-estimated mean Ki for physiological uptake (Ki-liver and Ki-spleen) and tumor lesions (Ki-tumor) using either IDIF or PBIF were also compared.
The mean AUC ratio (PBIF/IDIF) was 0.98 ± 0.06. The mean Ki bias between PBIF and IDIF was -2.6 ± 6.2% (confidence interval, CI: -5.8; 0.6). For Ki-spleen and Ki-tumor, low relative bias with low SD were found [4.65 ± 7.59% (CI: 0.26; 9.03) and 3.70 ± 8.29% (CI: -1.09; 8.49) respectively]. For Ki-liver analysis, relative bias and SD were slightly higher [7.43 ± 13.13% (CI: -0.15; 15.01)].
Our study showed that the PBIF approach allows for reduction in WBdyn DOTATOC-PET/CT acquisition times with a minimum gain of 20 min.
验证一种基于人群的输入函数（PBIF）模型，该模型可减少动态全身（WBdyn）PET自注射时间起的扫描需求。
纳入37例疑似/已知高分化神经内分泌肿瘤患者（GAPETNET试验NTC03576040）。所有WBdyn 68Ga-DOTATOC-PET/CT采集均在数字PET系统上进行（一次以心脏为中心的6分钟步长，随后进行九次全身扫描）。PBIF模型由20个图像衍生输入函数（IDIF）构建而成，这些函数通过自动左心室分割工具从相应数量患者的WBdyn检查中获得。所有IDIF峰值均与中位达峰时间对齐，根据患者体重和给药活性进行归一化，然后拟合到指数模型函数。然后通过缩放PBIF使其在注射后20 - 55分钟时间窗（对应全身扫描3 - 7次）内与相应的IDIF部分匹配，将其应用于17项独立患者研究。使用布兰德 - 奥特曼分析（平均偏差±标准差）比较IDIF和PBIF的曲线下面积（AUC）之比。还比较了使用IDIF或PBIF时，帕塔克估计的生理摄取（肝Ki和脾Ki）和肿瘤病变（肿瘤Ki）的平均Ki。
平均AUC比（PBIF/IDIF）为0.98±0.06。PBIF和IDIF之间的平均Ki偏差为 -2.6±6.2%（置信区间，CI： -5.8；0.6）。对于脾Ki和肿瘤Ki，发现相对偏差低且标准差低[分别为4.65±7.59%（CI：0.26；9.03）和3.70±8.29%（CI： -1.09；8.49）]。对于肝Ki分析，相对偏差和标准差略高[7.43±13.13%（CI： -0.15；15.01）]。
我们的研究表明，PBIF方法可减少WBdyn DOTATOC-PET/CT采集时间，最少可减少20分钟。
