variability and reliability of normal liver ^18f-fdg standardized uptake value in relation to placement of volume of interest

Background: the impact of volume-of-interest (voi) placement on liver standardized-uptake-value (suv) measurement remains unclear. objectives: to investigate variability and reliability of liver suv in relation to placement of voi within the right lobe of the liver. patients and methods: positron emission tomography-computed tomography (pet/ct) of 343 patients with normal liver were retrieved. liver mean suv (suvmean ) were measured by placing spherical vois at upper, portal vein and lower level of the right-lobe liver. the general linear model, bland-altman analysis, one-way anova were performed and the intraclass correlation coefficients (icc) were calculated to identify impacts of reader and voi position on variations of liver suvmean . results: the liver suvmean at the upper, lower and portal vein level were 1.85 ± 0.435, 1.88 ± 0.409, and 1.89 ± 0.414, respectively, with no significant inter-location difference (p = 0.449). the coefficients of variation were close among locations (21.9% - 23.5%). the inter-location iccs controlling for reader were 0.955 for reader 1 and 0.980 for reader 2, while the inter-reader iccs controlling for location were between 0.930 and 0.937, indicating great reliability and reproducibility. however, inter-reader bias of suvmean were significantly different among locations (p < 0.001). bias at portal vein level (0.069 ± 0.112) was significantly larger than those at the upper level (0.004 ± 0.144, p < 0.001) and lower level (0.002 ± 0.136, p < 0.001). conclusion: liver suvmean measured by placing voi at the portal vein level exhibited a relatively larger variation and less precision between readers, which was not recommended.