Influence of PET reconstruction technique and matrix size on qualitative and quantitative assessment of lung lesions on [18F]-FDG-PET: A prospective study in 37 cancer patients

PurposeTo evaluate the influence of point spread function (PSF)-based reconstruction and matrix size for PET on (1) lung lesion detection and (2) standardized uptake values (SUV).MethodsThis prospective study included oncological patients who underwent 18F]-FDG-PET/CT for staging. PET data were reconstructed with a 2D ordered subset expectation maximization (OSEM) algorithm, and a 2D PSF-based algorithm (TrueX), separately with two matrix sizes (168 × 168 and 336 × 336). The four PET reconstructions (TrueX-168; OSEM-168; TrueX-336; and OSEM-336) were read independently by two raters, and PET-positive lung lesions were recorded. Blinded to the PET findings, a third independent rater assessed lung lesions with diameters of >4 mm on CT. Subsequently, PET and CT were reviewed side-by side in consensus. Multi-factorial logistic regression analyses and two-way repeated measures analyses of variance (ANOVA) were performed.ResultsThirty-seven patients with 206 lung lesions were included. Lesion-based PET sensitivities differed significantly between reconstruction algorithms (P < 0.001) and between reconstruction matrices (P = 0.022). Sensitivities were 94.2% and 88.3% for TrueX-336; 88.3% and 85.9% for TrueX-168; 67.8% and 66.3% for OSEM-336; and 67.0% and 67.9% for OSEM-168; for rater 1 and rater 2, respectively. SUV_max and SUV_mean were significantly higher for images reconstructed with 336 × 336 matrices than for those reconstructed with 168 × 168 matrices (P < 0.001).ConclusionOur results demonstrate that PSF-based PET reconstruction, and, to a lesser degree, higher matrix size, improve detection of metabolically active lung lesions. However, PSF-based PET reconstructions and larger matrix sizes lead to higher SUVs, which may be a concern when PET data from different institutions are compared.