Reproducibility of SUVmax in serial PET studies

Objectives SUVs quantitatively measure tumor uptake and may be used to monitor tumor response.Here we assess reproducibility and uncertainty of SUV_max and establish minimum criteria for determining the significance of SUV_max changes in serial PET studies.

Methods A PET-ACR phantom background and spheres were filled with ¹⁸F 1-to-5 concentration ratio, and scanned for 5 hrs in list-mode. The data were parsed into 10 runs of equal total counts,corresponding to those obtained in a 4min clinical liver scan.3D ROIs were placed over each sphere and the background,and compared to determine the correlation between runs.The maximum activity pixel position in the spheres was calculated for all 10 runs to determine the statistical uncertainty of the SUVmax position.Two consecutive PET/CT scans were obtained for 10 SCC patients. Lesion ROIs were compared pixel-by-pixel; correlation coefficients and position uncertainties were determined.

Results Average correlation across 10 runs for each sphere was: 96.4±0.8% 96.4±0.6% 95±1%.The uniform background ROI average correlation was 95.6±0.6%.The correlation between image sets averaged over all the patients was 96.6±0.9%.The inter-scan change in position of SUVmax,averaged over all the patients,was 11.1±1.5mm (~ 3 pixels), and the average inter-run SUVmax difference was 6.5±2.5%.

Conclusions Inter-scan reproducibility is high for both phantom and patients at clinical scan times. However, there can be significant displacement and variation of the value of the SUVmax even between consecutive scans.This can have a significant impact when using SUVmax as a metric for determining treatment outcome,especially if the difference in SUV’s is less than 6%.