Reproducibility of uptake estimates in FDG PET: a Monte Carlo study

Tumor glucose metabolism measurements from Fluoro-deoxyglucose (FDG) Positron Emission Tomography (PET) could be helpful for therapeutic follow-up, assuming the index used is highly reproducible in the absence of change in tumor metabolism. We compared the variability of uptake values as estimated by Tumor-to-Normal tissue Ratio (TNR) depending on whether they are calculated from maximum value (TNRmax) or from the mean value (TNRavg) in a volume of interest (VOI). Monte Carlo simulations of CPET acquisitions using a data spectrum thorax phantom including eight spheres mimicking tumors with various sizes and locations were performed with SimSET. The phantom and acquisition protocols were designed to study the differences in TNRavg and TNRmax estimates when there were changes in tumor size, tumor-to-background activity contrast, tumor uptake, noise level in the image, or VOI location with respect to the tumor. Twenty realizations of each configuration were simulated. Both TNRavg and TNRmax were highly sensitive to a 50% change in tumor uptake. However, uptake estimates also differed by more than 25% when the tumor uptake did not change but the tumor size decreased by 30, 50 or 70%. With TNRavg, this dependency upon tumor size could be almost fully removed using partial volume effect correction. When only the non-specific activity changed, TNRmax could vary by as much as 50% for small spheres, while TNRavg variations remained less than 25%. TNRmax did not depend on the VOI location with respect to the tumor, while the reproducibility of TNRavg was highly dependent on the correct positioning of the VOI with respect to the tumor. TNRavg and TNRmax were both sensitive to the noise level in the images, suggesting that repeated scans should be performed with similar counting rate and duration. Overall, TNRavg appears less sensitive to changes different from tumor uptake variations and should be preferred to TNRmax in the context of therapeutic follow-up, provided a correcti- - on for partial volume effect is used and the positioning of the VOI with respect to the tumor is precisely controlled