Abstract
BACKGROUND: Although quantitative analysis using standardized uptake value (SUV) becomes realistic in clinical single-photon emission computed tomography/computed tomography (SPECT/CT) imaging, reconstruction parameter settings can deliver different quantitative results among different SPECT/CT systems. This study aims to propose a use of the digital reference object (DRO), which is a National Electrical Manufacturers Association (NEMA) phantom-like object developed by the Quantitative Imaging Biomarker Alliance (QIBA) fluorodeoxyglucose-positron emission tomography technical committee, for the purpose of harmonizing SUVs in Tc-99m SPECT/CT imaging. METHODS: The NEMA body phantom with determined Tc-99m concentration was scanned with the four state-of-the-art SPECT/CT systems. SPECT data were reconstructed using different numbers of the product of subset and iteration numbers (SI) and the width of 3D Gaussian filter (3DGF). The mean (SUV(mean)), maximal (SUV(max)), and peak (SUV(peak)) SUVs for six hot spheres (10, 13, 17, 22, 28, and 37 mm) were measured after converting SPECT count into SUV using Becquerel calibration factor. DRO smoothed by 3DGF with a FWHM of 17 mm (DRO(17 mm)) was generated, and the corresponding SUVs were measured. The reconstruction condition to yield the lowest root mean square error (RMSE) of SUV(means) for all the spheres between DRO(17 mm) and actual phantom images was determined as the harmonized condition for each SPECT/CT scanner. Then, inter-scanner variability in all quantitative metrics was measured before (i.e., according to the manufacturers' recommendation or the policies of their own departments) and after harmonization. RESULTS: RMSE was lowest in the following reconstruction conditions: SI of 100 and 3DGF of 13 mm for Brightview XCT, SI of 160 and 3DGF of 3 pixels for Discovery NM/CT, SI of 60 and 3DGF of 2 pixels for Infinia, and SI of 140 and 3DGF of 15 mm for Symbia. In pre-harmonized conditions, coefficient of variations (COVs) among the SPECT/CT systems were greater than 10% for all quantitative metrics in three of the spheres, SUV(max) and SUV(mean), in one of the spheres. In contrast, all metrics except SUV(max) in the 17-mm sphere yielded less than 10% of COVs after harmonization. CONCLUSIONS: Our proposed method clearly reduced inter-scanner variability in SUVs. A digital phantom developed by QIBA would be useful for harmonizing SUVs in multicenter trials using SPECT/CT.