Phantoms
Catphan 500 and 600 (The Phantom Laboratory, Salem, NY, USA) designed for assessing various CT image quality parameters were used in this study [1]. The phantoms consist of 4/5 test modules enclosed in 20 cm housing. In this study, sensitometry (CTP401/CTP404), image uniformity (CTP486), low contrast (CTP515) and high resolution (CTP528) modules were used. The sensitometry module in Catphan 500 (CTP401) consist of air, teflon, acrylic, and low-density polyethylene (LDPE) inserts [1]. In addition to the above-mentioned inserts, the CTP404 module in Catphan 600 has delrin, polystyrene and polymethylpentene (PMP) [1]. Theoretical Hounsfield unit (HU) of these materials at 120 kVp are: 990 HU (teflon), 340 HU (delrin), 120 HU (acrylic), -35 HU (polystyrene), -100 HU (LDPE), -200 HU (PMP) and -1000 HU (air ) [1].
CT scanners
Image quality testing was performed on seven CT scanners of different models from four vendors (Table 1 ), in the same hospital.
Acquisition protocols
The phantoms were scanned repeatedly using the same acquisition protocol over a four year period (2016 – 2019) for all scanners; one helical scan covering all the modules in the phantom at 120 kVp, ~ 20 mGy CTDIvol, 1 s rotation time, ~ 1 pitch and the most used collimation on each scanner. All images were reconstructed with the standard soft-tissue kernel, 210 mm displayed field of view, 2.5/3 mm slice thickness as well as the thinnest slice available on each scanner.
Image quality evaluation
All image quality parameters were evaluated automatically using the ImageOwl software (Fig. 1 , Image Owl, Inc. Greenwich, NY, USA). CT number of each sensitometry insert was estimated by calculating the mean HU in a circular ROI placed in each insert ( Fig. 1A). Linearity was estimated by finding the correlation between the measured CT number and the linear attenuation coefficients of each material at 60 keV[1].
Slice thickness (defined as the full width at half maximum (FWHM) of the slice sensitivity profile) was evaluated by making a Gaussian fit to the wires angled at 23° in the sensitometry module (Fig. 1B). FWHM was then estimated and multiplied by 0.42 to find the slice thickness. Noise was measured as the standard deviation of CT numbers in a ROI placed in the center of the uniformity module covering 40% of the diameter (Fig. 1C). Uniformity was estimated as the absolute difference between the mean CT number in central ROI and the upper, lower, right and left ROIs (Fig. 1D).
Low contrast resolution was assessed by detecting the smallest size of super-slice targets with different contrast (0.3, 0.5 and 1%), and diameters (2 – 15 mm, Fig. 1E). The beads located in the high resolution module were used for the evaluation of critical frequencies (50, 10, 5 and 2%) of the modulation transfer function (MTF, Fig. 1F). For all tests, values obtained from the acceptance testing were used as baselines whereas tolerance levels were set according to the recommendations from the Intentional Electrotechnical Commission (IEC, [2, 3]), the American Association of Physicists in Medicine (AAPM, [4]) and Institute of Physics and Engineering in Medicine (IPEM,[5] ).
