Comparison of different automated lesion delineation methods for metabolic tumor volume of 18F-FDG PET/CT in patients with stage I lung adenocarcinoma

The aim of the study was to investigate the suitable segmentation method in small, low uptake and heterogeneous nodules of stage I lung adenocarcinoma.133 stage I lung adenocarcinoma patients with F-FDG PET/CT scans were enrolled in this retrospective study. All lesions were divided into different groups according to nodule density, nodule size, and the maximum standard uptake value (SUVmax) level. Four different PET segmentation methods were performed, including percentage threshold of SUVmax (T42% and T42% × RC), gradient-based threshold (adaptive iterative algorithm, AT-AIA), and background-related threshold (adaptive thresholding at 40% SUVmax, AT40%) approaches. The MTVs were evaluated and compared with CT volume (CTV). Percentage volume error (%VE) compared to CTV was calculated and the correlations between MTVs and CTV were analyzed.AT-AIA had the highest accuracy in large, high uptake, and solid nodules (72.5%, 72.4%, and 65.6%, respectively). AT40% had the highest accuracy in small, low uptake and nonsolid nodules (56.6%, 56.1%, and 62.6%, respectively). In part-solid nodules, the accuracy of AT-AIA (60.0%) and AT40% (56.7%) were higher than that of T42% and T42% × RC. The MTV of AT-AIA was in excellent correlation with the CTV in solid nodules (R = 0.831, P < .001) and in high uptake nodules (R = 0.830, P < .001). The MTV of AT40% was in good correlation with the CTV in nonsolid nodules (R = 0.686, P = .003) and in part-solid nodules (R = 0.731, P < .001).AT40% showed best performance in small, low uptake, nonsolid and part-solid lesions. AT-AIA was suitable for large, high uptake, and solid lesions.