Repeatability of ^sup 18^F-FDG PET/CT total metabolic active tumor volume and total tumor burden assessment in NSCLC patients: impact of acquisition, image reconstruction and delineation method

Objectives: In this study we investigated the effects of uptake time, acquisition protocol, image reconstruction settings and delineation methods on total metabolic active tumor volume (MATV) and total tumor burden (TTB = MATV × SUVmean summed across all lesions) and their repeatability for 18F-FDG PET/CT studies in non-small cell lung cancer (NSCLC) patients. Methods: Ten stage III and IV NSCLC patients underwent 2 baseline 18F-FDG PET/CT studies on separate days. At each time point, scans were obtained at both 60 and 90 minutes post injection. All PET/CT data were reconstructed using EARL compliant (EARL) and resolution modeling with point spread function (PSF) protocols. The reconstructed images were assessed with four semi-automated VOI segmentation methods: isocontours at SUV = 2.5, SUV = 4.0, 41% of SUVmax and a contrast adapted 50% of SUVpeak method. In addition, a consensus VOI was derived on the agreement across the predefined VOIs based on the majority vote method, i.e., when a voxel was selected by 2 or more of those VOIs (MV2). The total MATV and TTB and their repeatability were then calculated for each reconstructed dataset using all of the 5 methods. The results were evaluated using Pearson’s correlations, box-plots and Bland-Altman analyses. The obtained delineation of all cases were visually inspected (by an experienced observer) and qualitatively ranked into four classes: success, under or overestimation of volume and failure. Results: The best success rate was obtained using MV2 at 60min tracer uptake with success on 90% cases, independent of reconstruction settings. For those scans, there was perfect correlation between EARL and PSF derived MATV (r² = 1.00, p = 0.92) at a mean MATV difference of -4.49% and repeatability coefficient (RC = 1.96 × SD) of 20.95%. The MATV test-retest (TRT) mean difference was 0.35% (RC = 24.07%) and -3.73% (RC = 20.39%) for EARL and PSF, respectively. Yet, best repeatability was obtained with 90min uptake scans (with a success on 88% of cases, independent of reconstruction settings), having a mean MATV TRT difference of 1.15% (RC = 6.78%) for EARL and 3.38% (RC = 15.57%) for PSF. The same trend was seen for TTB, with repeatability coefficients of 11.88% for EARL and 16.66% for PSF. Most importantly, MV2 was never the worst segmentation, independent of tracer uptake time and image reconstruction settings (total success of 89%), while 11% of the cases showed an overestimated MATV. The performance of MV2 was no