The phantom used was a Jaszczak phantom (Deluxe ECT Data Spectrum Corporation, Chapel Hill, USA), The phantom was a standardized diameter 21.6cm. Four capillary line sources made of plastic material with an inner diameter of 1mm were imbedded parallel to the cylindrical shell of the trunk in water to assess tomographic resolution with scatter. The line sources, filled with 99mTc(37MBq/cm) were positioned at the center, at 2.5cm, at 5 cm and at 7.5cm away from the center of the phantom, depicted in Fig.1(computed-tomography image), these positions were chosen to permit assessment of the spatial resolution at different distances. Phantom included six cold spheres (diameters respectively:9.5,12.7,15.9,19.1,25.4,31.8 mm) and a cold rod insert (diameters respectively:4.8,6.4,7.9,9.5,11.1,12.7 mm) was adopted to evaluate image quality with the background activity of 370MBq 99mTc. Image acquisition was started 2 hours later, depicted in Fig.2.
Image acquisition was performed with CZT-SPECT/CT (Discovery NM/CT 670 CZT; GE Healthcare), equipped with a wide energy high resolution collimator (WEHR) and conventional SPECT/CT (Discovery NM/CT 670; GE Healthcare), equipped with a low energy high resolution collimator (LEHR). Necessary quality control is conducted periodically to maintain the image quality in daily work. The specified phantom shall be aligned with the system's axis of rotation and centered in the field of view. SPECT images were obtained in step-and-shoot mode, the acquisitions were performed using a 180° non circular orbit for each detector, 3° per projection and 120 projection angles in list mode, 256×256 matrix size (pixel size of 2.21 mm), zoom = 1, body contour mode was adopted in the acquisition, total count per projections collected of the line sources is not less than 100k according to NEMA(1), Image quality phantom with inserts was separately collected images 10s, 20s, 30s per projections to evaluate the effect of acquisition time on image quality. all acquisitions were performed with an energy window at 20% centered at 99mTc photopeak (140 keV) of conventional SPECT/CT and 15% of CZT-SPECT/CT. A CT scan was performed with parameters 120 kV, 220 mA after the SPECT acquisition. The CT data were used for attenuation correction. Dual energy window scatter correction technique (120Kev±5%) was used for two cameras.
Image reconstruction and data analysis
All reconstructions were performed using Xeleris workstation, the reconstruction algorithms considered in this study were FBP and OSEM algorithm, for each reconstruction parameters, the FWHM values were measured.
Varying reconstruction parameters in OSEM of line sources reconstruction were as following: the number of subsets was constantly set as 10, (1) the number of iterations was varied from 1 to 20(1,2,3,4,5,6,7,8,9,10,15,20 respectively) with a 4mm Gauss filter. (2) Based on (1), the number of iterations was selected and fixed, Butterworth filter was applied, and the cut-off frequency of Butterworth filter ranged from 0.40 to 0.90(0.40,0.50,0.60,0.70,0.80,0.90 cycles per pixel, respectively) with order being set to 10. (3) Based on (1), the number of iterations was fixed, Gauss filter was applied (characterized by full width at half-maximum values, FWHM), and the Gauss filter ranged from 1.00 to 6.00(FWHM=1.00,2.00,3.00,4.00,5.00,6.00mm, respectively). (4) no-filter was applied and the number of iterations was varied from 1 to 20(1,3,5,7,9,15,20, respectively). CTAC, SC and RR were adopted in (1)-(4). (5) Based on (1), the number of iterations was fixed with a 4mm Gauss filter, AC, SC, RR algorithms and no correction (NC) were adopted separately to evaluate the influence of the correction algorithms on the spatial resolution.
FBP algorithm with Butterworth filter was also performed to reconstruct tomographic images of line sources, with the cut-off frequency ranging from 0.4 to 0.9(0.4,0.5,0.6,0.7,0.8,0.9 cycles per pixel, respectively), order being set to 10. CTAC was adopted by default.
Image quality phantom with inserts (acquisition time of 10s,20s,30s per projections) was reconstructed by OSEM with a 4mm Gauss filter, the number of iterations was fixed based on (1), and the number of subsets was fixed at 10, CTAC, SC and RR algorithms were adopted.
Spatial resolution: According to the NEMA, we reconstructed transverse slice through center of line sources and thickness of reconstructed slice was 8.84mm (10±3mm recommended by NEMA), four reconstructed point sources in the reconstructed slices shall be analyzed individually with a square ROI (regions of interest). Each ROI shall be centered on the maximum count pixel. The size of this square ROI must be at least four times the FWHM of the count profile to be measured. For each point source in the images, the FWHM in radial and tangential shall be determined(1). Different to the NEMA standard, we measured on the center slice of line sources only. Measured spatial resolution values were plotted as FWHM (mm) values with varying reconstruction parameters.