A threshold method to improve standardized uptake value reproducibility

Although standardized uptake values (SUV) are widely used to quantify the uptake of 18F-fluorodeoxyglucose (18F-FDG) in tumours, there are systematic differences in the way this index is applied by different investigators. The aims of this study were to compare the effects of using maximum or mean region counts in the calculation of SUV and to investigate an alternative technique based on a fixed fraction of the maximum counts. Simulated PET projections of the thorax were generated together with spherical lesions that varied in diameter from 1.6 to 4.8 cm with uptake values of 2, 4 and 8. The lesion SUVs were determined using either the maximum (SUVmax) or mean count (SUVmean) values found in regions circumscribing the lesion. In addition, average values were calculated that only included region pixels that exceeded a selected fraction of maximum value (SUV0.6max or SUV0.75max). These methods were also applied to six clinical 18F-FDG PET studies with a total of 12 lesions. The SUVs for these lesions were determined independently by four observers. Decreases with respect to SUVmax of 57%, 23% and 14% were found for SUVmean, SUV0.6max and SUV0.75max approaches respectively in the simulation study. The variation in SUVmean with region size was 35%, while the SUV0.6max and SUV0.75max was less than 3%. Similar results were obtained for the clinical data. We conclude that the proposed technique produces SUVs that are essentially independent of lesion region size and shape. It is expected that this will provide a more stable and reliable result than current approaches.