Kolinger Guilherme D, Vállez García David, Kramer Gerbrand M, Frings Virginie, Smit Egbert F, de Langen Adrianus J, Dierckx Rudi A J O, Hoekstra Otto S, Boellaard Ronald
University of Groningen, University Medical Center Groningen, Department of Nuclear Medicine and Molecular Imaging, Hanzeplein 1, 9713 GZ, Groningen, The Netherlands.
Amsterdam University Medical Centers, location VU Medical Center, Department of Radiology and Nuclear Medicine, De Boelelaan 1117, 1081 HV, Amsterdam, The Netherlands.
EJNMMI Res. 2019 Feb 7;9(1):14. doi: 10.1186/s13550-019-0481-1.
Total metabolic active tumour volume (TMATV) and total tumour burden (TTB) are increasingly studied as prognostic and predictive factors in non-small cell lung cancer (NSCLC) patients. In this study, we investigated the repeatability of TMATV and TTB as function of uptake interval, positron emission tomography/computed tomography (PET/CT) image reconstruction settings, and lesion delineation method. We used six lesion delineation methods, four direct PET image-derived delineations and two based on a majority vote approach, i.e. intersection between two or more delineations (MV2) and between three or more delineations (MV3). To evaluate the accuracy of those methods, they were compared with a reference delineation obtained from the consensus of the segmentations performed by three experienced observers. Ten NSCLC patients underwent two baseline whole-body [F]2-Fluoro-2-deoxy-2-D-glucose ([F]FDG) PET/CT studies on separate days, within 3 days. Two scans were obtained on each day at 60 and 90 min post-injection to assess the influence of tracer uptake interval. PET/CT images were reconstructed following the European Association of Nuclear Medicine Research Ltd. (EARL) compliant settings and with point-spread-function (PSF) modelling. Repeatability between the measurements of each day was determined and the influence of uptake interval, reconstruction settings, and lesion delineation method was assessed using the generalized estimating equations model.
Based on the Jaccard index with the reference delineation, the MV2 lesion delineation method was the most successful method for automated lesion segmentation. The best overall repeatability (lowest repeatability coefficient, RC) was found for TTB from 90 min of tracer uptake scans reconstructed with EARL compliant settings and delineated with 41% of lesion's maximum SUV method (RC = 11%). In most cases, TMATV and TTB repeatability were not significantly affected by changes in tracer uptake time or reconstruction settings. However, some lesion delineation methods had significantly different repeatability when applied to the same images.
This study suggests that under some circumstances TMATV and TTB repeatability are significantly affected by the lesion delineation method used. Performing the delineation with a majority vote approach improves reliability and does not hamper repeatability, regardless of acquisition and reconstruction settings. It is therefore concluded that by using a majority vote based tumour segmentation approach, TMATV and TTB in NSCLC patients can be measured with high reliability and precision.
全代谢活性肿瘤体积（TMATV）和肿瘤总负荷（TTB）作为非小细胞肺癌（NSCLC）患者的预后和预测因素，受到越来越多的研究。在本研究中，我们调查了TMATV和TTB作为摄取间隔、正电子发射断层扫描/计算机断层扫描（PET/CT）图像重建设置以及病变勾画方法的函数的可重复性。我们使用了六种病变勾画方法，四种直接从PET图像得出的勾画方法以及两种基于多数投票法的方法，即两种或更多勾画之间的交集（MV2）以及三种或更多勾画之间的交集（MV3）。为了评估这些方法的准确性，将它们与由三名经验丰富的观察者进行的分割共识得出的参考勾画进行了比较。十名NSCLC患者在3天内于不同日期接受了两次基线全身[F]2-氟-2-脱氧-D-葡萄糖（[F]FDG）PET/CT检查。每天在注射后60分钟和90分钟各进行一次扫描，以评估示踪剂摄取间隔的影响。PET/CT图像按照欧洲核医学研究协会（EARL）的合规设置并采用点扩散函数（PSF）建模进行重建。确定每天测量之间的可重复性，并使用广义估计方程模型评估摄取间隔、重建设置和病变勾画方法的影响。
基于与参考勾画的杰卡德指数，MV2病变勾画方法是自动病变分割最成功的方法。在按照EARL合规设置重建并使用病变最大SUV的41%方法勾画的示踪剂摄取90分钟扫描中，TTB的总体可重复性最佳（可重复性系数，RC最低）（RC = 11%）。在大多数情况下，TMATV和TTB的可重复性不受示踪剂摄取时间或重建设置变化的显著影响。然而，一些病变勾画方法应用于相同图像时，其可重复性存在显著差异。
本研究表明，在某些情况下，TMATV和TTB的可重复性会受到所使用的病变勾画方法的显著影响。无论采集和重建设置如何，采用多数投票法进行勾画可提高可靠性且不妨碍可重复性。因此得出结论，通过使用基于多数投票的肿瘤分割方法，可以高度可靠且精确地测量NSCLC患者的TMATV和TTB。
