Assessment of factors affecting accuracy of standardised uptake values in positron emission tomography

[摘要] English: Positron emission tomography (PET) is an imaging method that uses tracers labelled withpositron emitting isotopes for the monitoring and evaluation of in vivo molecularprocesses. Semi-quantitative determination of tracer uptake in a lesion is accomplishedby calculating the standardised uptake value (SUV), an index that represents the amountof uptake in a given volume-of-interest (VoI) in relation to the average uptake throughoutthe body. The SUV is influenced by biological and physical factors that determine theuptake or detectability of the tracers which may result in false results. Changes in SUVof small lesions or lesions with low activity uptake cannot be determined with enoughcertainty and precision to be used for decision-making and it is therefore necessary toinvestigate the factors affecting the SUV.The aim of this study was to assess the relative importance of the physical factors thataffect the accuracy of a single SUV measurement using Monte Carlo modelling.Phantom studies were performed to determine the influence of the partial volume effectdue to spatial resolution using a PET scanner. Comparative Monte Carlo simulationswere performed on a computer cluster using a voxelised version of the same phantom.The XCAT anthropomorphic phantom was used to assess the influences on SUV in ahuman-like configuration and was set-up to simulate movement in the thorax duringbreathing. SUVs were calculated using simulations of the phantom in 2D and 3D modesto assess the influence of the partial volume effect by variation of the size of the lesions,by variation of the contrast ratios and by placing the lesions in different areas in the lungsduring. Influence of activity from outside the field-of-view (FoV) was also assessed aswell as the impact the various coincidence types have. Statistical methods were used tocompare the difference in data for statistical significance.It was found that the partial volume effect was present when evaluating the SUVs of theactivity in the spheres of the phantom when scanned on a PET/CT scanner as well aswhen performing Monte Carlo simulations. Statistically there were no significant differences between the two scanning modes. The mean SUV increased as the voxelsizes became smaller. The choice of matrix influenced the amount of partial volumeeffect.The relative contributions of true-, scatter- and random coincidences demonstrated thatthe true coincidences were the major contributor when assessing the data from thisphantom. The relative contribution of the trues-to-total coincidences decreased with adecrease in lesion size and contrast ratio whereas the relative contributions of thescattered- and random coincidences increased. The contributions of scatters and randomsincreased during the 3D acquisition mode compared to 2D mode. The contribution of thetrues-to-total coincidences decreased with an increase in VoI size and consequentlycaused a decrease in the mean SUV.The location of the lesion made a difference in SUV when the same size lesions arecompared to each other. Apical lesions experienced the least amount of motion duringbreathing, were distorted less and had the least amount of variation in SUV. By movingthe phantom partly outside the FoV, significant effects on the SUVs of objects still insidethe FoV were found. An increase in the SUVs was observed when the true coincidenceswere used for the calculation. A decrease in true SUVs was found at the right basallesion.In conclusion, partial volume effects play a significant role when determining the SUV ofobjects based on their size and contrast ratio; the location of pulmonary lesions affectsSUV calculation during breathing; and activity outside the field-of-view of the scannercontributed to a change in SUV in particular to the central and basal regions of the lung.

[发布机构] University of the Free State