Phantoms
The 20-cm diameter American College of Radiology (ACR) Quality Assurance phantom (Gammex 464, Gammex, Middleton, USA) was scanned to assess the image quality by measuring the noise power spectrum (NPS) and task-based transfer function (TTF). This phantom was placed in its elliptical ring (26 × 33 × 16 cm3) to simulate the abdomen (Fig. 1a). A multienergy CT phantom (Multi-Energy CT phantom, Sun Nuclear, Middleton, USA) associated with its elliptical body insert (30 × 40 × 15 cm3) was also used to assess the contrast between iodine and solid water inserts (Fig. 1b).
CT scanners and scanning protocols
All acquisitions were performed on the Aquilion ONE PRISM Edition CT system (Canon Medical Systems, Otawara, Japan). Both phantoms were scanned three times with the DECT mode using the same clinical protocol for abdomino-pelvic examinations: a 80/135 kVp switching, a collimation of 80 × 0.5 mm, a rotation time of 1 s and a pitch value of 0.813. The tube current was set at 230 mA to obtain a CT volume dose index of 12.6 mGy, close to the French national diagnostic reference level for abdomen and pelvis examinations fixed at 13 mGy in France [34].
Raw data were reconstructed using the two versions of the DLSR (DLSR V1 and DLSR V2), the “Body spectral” reconstruction kernel, and using the three available DLSR levels: mild, standard, and strong. As the DLSR did not allow a 1-mm thick reconstruction, all raw data were reconstructed using the 0.5-mm slice thickness and a 0.5-mm increment for both DLSR versions. The field of view used was 250 mm for the ACR phantom, and 420 mm for the multienergy CT phantom.
For each acquisition, VMIs were reconstructed using the Vitrea workstation (Canon Medical Informatics, Minnetonka, Minnesota, USA) at four low energy levels (40, 50, 60, and 70 keV) used in clinical practice to improve the iodine contrast.
Assessment of iodine contrast on VMIs
For each DLSR level and for both DLSR versions, the HU values were measured in the central slice of the multienergy CT phantom. One circular region of interest (ROI) of 2-cm diameter was placed on the solid water insert and on three iodine inserts with an iodine concentration of 2, 1 and 0.5 mg/mL (Fig. 2a). For each insert, the mean HU value within each ROI was computed for the VMIs at 40, 50, 60, and 70 keV.
The contrast between the iodine and solid water inserts was calculated at each energy level for all DLSR levels and for both DLSR versions according to the following formula:
$$Contrast={HU}_{iodine}-{HU}_{solid\ water}\kern0.5em$$
(1)
where, HUiodinecorresponds to the mean HU value of each iodine insert (0.5, 1.0, and 2.0 mg/mL) and HUsolid water to the solid water value.
Task-based image quality assessment on VMIs
A task-based image quality assessment was performed using the ImQuest software (version 7.1, Duke University, USA) to assess the noise magnitude and texture using the NPS and the spatial resolution using the TTF [35, 36]. The detectability index (d′) was computed to assess the ability of the radiologist to detect enhanced lesions. All these metrics were calculated for three DLSR levels for the two versions, and for all VMI levels (40, 50, 60, and 70 keV).
Noise power spectrum
For each DLSR level and for both spectral versions, the NPS was computed on the uniform module of the ACR phantom using 40 consecutive slices. Four square ROIs of 128 x 128 pixels were placed in this uniform module (Fig. 2b) and the NPS was calculated following this formula:
$${NPS}_{2D}\left({f}_x,{f}_y\right)=\frac{\Delta _x{\Delta }_y}{L_x{L}_y}\frac{1}{N_{ROI}}\sum_{i=1}^{N_{ROI}}{\left|{FFT}_{2D}\left\{{ROI}_i\left(x,y\right)-{FIT}_i\left(x,y\right)\right\}\right|}^2$$
(2)
where Δx and Δy are the pixel sizes in the x- and y-directions; FFT is the Fast Fourier Transform; Lx and Ly are the lengths of the ROIs in the x- and y-directions; NROI is the number of ROIs; ROIi(x, y) is the mean pixel value measured for a ROI at the position (x, y) and FITi(x, y) is a second order polynomial fit of ROIi(x, y). The noise magnitude and the average spatial frequency (fav) were calculated to quantify the noise level and noise texture respectively. A fav at low spatial frequencies may indicate a blotchy noise appearance. The following formula was used to compute the fav values:
$${f}_{av}=\frac{\int f. NPS(f) df}{\int NPS(f) df}$$
(3)
where f is the radial spatial frequency and NPS(f) is the radially re-binned/average 1D NPS [36].
Task-based transfer function
The TTF was assessed using the acrylic insert of the ACR phantom (Fig. 2c) following the methodology reported by Richard et al. [37]. A circular ROI was placed around the insert and a circular-edge technique was used to measure the edge spread function, which was obtained by calculating the radius of each pixel from the centre of each pixel of the insert. The line spread function was obtained by derivation of the edge spread function. The TTF was then computed from the line spread function normalised Fourier transformation. It was computed using 20 consecutive slices.
Detectability index
Three task functions were defined to model the detection of enhanced lesions of 10 mm diameter with low iodine concentrations of 0.5, 1, and 2 mg/mL. The TTF results for the acrylic insert were used for each detection task combined with the NPS to calculate the detectability index (d′) using a non-prewhitening observer model with an eye filter [38]:
$${d^{\prime}}_{NPWE}^2=\frac{{\left[\iint \left|W\left(u,v\right)\right|2. TTF{\left(u,v\right)}^2.E{\left(u,v\right)}^2 dudv\right]}^{{}^2}}{\iint {\left|W\left(u,v\right)\right|}^2. TTF{\left(u,v\right)}^2. NPS{\left(u,v\right)}^2.E{\left(u,v\right)}^4 dudv}$$
(4)
where u and v are the spatial frequencies in the x- and y-directions, E the eye filter that models the human visual system sensitivity to different spatial frequencies, and W(u,v) the task function defined as:
$$W=\left|F\left\{{h}_1\left(x,y\right)-{h}_2\left(x,y\right)\right\}\right|$$
(5)
where h1(x, y) and h2(x, y) correspond to the object present and the object absent hypotheses.
The contrast of each clinical task was measured directly on the iodine inserts of the multienergy CT phantom for each corresponding iodine concentration and for each energy level. The reading conditions used to obtain d’ were a 1.5 zoom factor, a viewing distance of 500 mm, a 300-mm field of view and a 0.05-mm pixel size.
Statistical analysis
Data are given as means and standard deviations. All quantitative data were compared between DLSR V1 and DLSR V2 using the Wilcoxon test for appeared samples. A p-value lower than 0.05 was considered significant.