Variation and repeatability of measured standardized uptake values depending on actual values: a phantom study.

Standardized uptake values (SUVs) are the most widely used quantitative imaging biomarkers in positron emission tomography (PET); however, little is known about the changes in variation and repeatability of SUVs depending on the magnitude of the values. We hypothesized that low SUVs have larger variations than high SUVs, and attempted various kinds of experimental PET scans using a phantom. By adjusting the ratio of F-18 solution to tap water, a NEMA IEC body phantom was set for SUVs of 2.0, 4.0, and 8.0 inside six hot spheres. PET data were obtained for 4 hours, and the data reconstructed every 2 min. The SUVmax and SUVpeak of the spheres in all images were recorded. The relative SUVs were calculated by dividing the measured SUV by actual SUV, and used for the Bland-Altman plots. Some variation was observed for the measured SUVs. The measured SUVs for the actual SUV of 2.0 showed the largest variation among those of 2.0, 4.0, and 8.0, and those of 8.0 showed the smallest. Similarly, the relative SUVs showed significantly larger variations for lower values. In addition, the relative SUVmax showed larger variation and value than the relative SUVpeak. The Bland-Altman plots showed considerable variation and little agreement, but the degree of variation decreased as the measured value increased. We demonstrated some variation of the measured SUVs, which decreased for larger measured values. Clinicians should consider the inaccuracy of low SUVs not only in daily practice, but also for multi-institutional studies.