Objectives: To assess variability of the average standard uptake value (SUV) computed by varying the number of hottest voxels within an (18)F-fluorodeoxyglucose ((18)F-FDG)-positive lesion. This SUV metric was compared with the maximal SUV (SUV(max): the hottest voxel) and peak SUV (SUV(peak): SUV(max) and its 26 neighbouring voxels).
Methods: Twelve lung cancer patients (20 lesions) were analysed using PET dynamic acquisition involving ten successive 2.5-min frames. In each frame and lesion, average SUV obtained from the N = 5, 10, 15, 20, 25 or 30 hottest voxels (SUV(max-N)), SUV(max) and SUV(peak) were assessed. The relative standard deviations (SDrs) from ten frames were calculated for each SUV metric and lesion, yielding the mean relative SD from 20 lesions for each SUV metric (SDr(N), SDr(max) and SDr(peak)), and hence relative measurement error and repeatability (MEr-R).
Results: For each N, SDr(N) was significantly lower than SDr(max) and SDr(peak). SDr(N) correlated strongly with N: 6.471 × N(-0.103) (r = 0.994; P < 0.01). MEr-R of SUV(max-30) was 8.94-12.63% (95% CL), versus 13.86-19.59% and 13.41-18.95% for SUV(max) and SUV(peak) respectively.
Conclusions: Variability of SUV(max-N) is significantly lower than for SUV(max) and SUV(peak). Further prospective studies should be performed to determine the optimal total hottest volume, as voxel volume may depend on the PET system.
Key points: • PET imaging provides functional parameters of (18) F-FDG-positive lesions, such as SUVmax and SUVpeak. • Averaging SUV from several hottest voxels (SUVmax-N) is a further SUV metric. • Variability of SUVmax-N is significantly lower than SUVmax and SUVpeak variability. • SUVmax-N should improve SUV accuracy for predicting outcome or assessing treatment response. • An optimal total hottest volume should be determined through further prospective studies.