Analysis of Jaszczak Phantom Measurement for Qualitative Evaluation of Image Resolution
Having adjusted the reconstruction time to the radionuclide-specific half-life, positron fraction (Table 1) and to the activity concentration in the phantom at the timepoint of imaging (Table 2), the total number of true events detected by the PET/CT scanner was similar in 18F-FDG (86.570.998 counts), 68Ga-HCl (81.659.978 counts) and 64Cu-HCl measurements (85.815.118 counts), respectively.
Figure 1 compares the resolution of PET images of the Jaszczak phantom filled with 18F-FDG, 68Ga-HCl and 64Cu-HCl, respectively. The minimum resolvable rods were those separated by 4.8 mm for 18F and 64Cu (smallest rods), respectively, and 6.4 mm for 68Ga measurements (second smallest rods). The smallest of the six cold spheres featuring a diameter of 9.5 mm was clearly recognizable in PET images of all three nuclides.
Analysis of NEMA PET Body Phantom Measurement for Quantitative Evaluation of Spatial Resolution
The number of true counts detected by the PET/CT scanner was similar for scans of the NEMA PET Body phantom filled with the three different nuclides (Table 2). Wrapping the phantom in cooling packs to simulate attenuation and scatter conditions similar to those in an obese patient reduced the number of detected true events by 36.1% (68Ga, 8:1 contrast ratio) to 38.4% (68Ga, 4:1 contrast ratio) (Table 2).
When comparing scans of the NEMA PET Body phantom at similar sphere-to-background contrast ratios and weight setup, spatial resolution was comparable between 18F and 64Cu measurements, but worse in the respective 68Ga measurement (Table 3). For each nuclide, the scan of the phantom mimicking an obese weight setup featured a slightly worse resolution compared to the scan of the phantom not wrapped with cooling packs. Resolution was slightly better in the PET image of phantom filled with a sphere-to-background activity concentration ratio of 8:1 compared to 4:1, especially in 68Ga-HCl PET/CT images.
Table 3
Spatial resolution (FWHM) in mm determined using the NEMA PET Body phantom.
|
4:1 contrast;
normal weight setup
|
4:1 contrast;
obese setup
|
8:1 contrast;
normal weight setup
|
8:1 contrast;
obese setup
|
18F-FDG
|
4.55 ± 0.18
|
4.83 ± 0.32
|
4.10 ± 0.21
|
4.21 ± 0.24
|
68Ga-HCl
|
5.35 ± 0.19
|
5.35 ± 0.33
|
4.83 ± 0.16
|
4.87 ± 0.24
|
64Cu-HCl
|
4.35 ± 0.20
|
4.55 ± 0.19
|
4.17 ± 0.30
|
4.30 ± 0.20
|
Legend Table 3: Spatial resolution is represented by mean and standard deviation of the Full Width at Half Maximum (FWHM) of all six spheres (in mm) of the NEMA PET Body phantom filled with 18F-FDG, 68Ga-HCl and 64Cu-HCl, respectively, at sphere-to-background activity concentration ratios of 4:1 and 8:1, respectively, and with applying cooling packs around the phantom (obese setup) or without (normal weight setup). |
Quantitative Evaluation of Image Quality
Representative for all NEMA PET Body phantom measurements, figure 2 compares RCmean, RCmax, RCpeak, CRC and CNR of all spheres of different sizes between 18F-FDG, 68Ga-HCl and 64Cu-HCl NEMA PET Body phantom measurements at a sphere-to-background activity concentration ratio of 8:1 and without the simulation of additional attenuation and scattering (obese setup). Mean values are compared between all NEMA PET Body phantom measurements in Attachment - Table 1A.
The mean activity concentrations recovered from the PET/CT images were comparable with the true activity concentrations in all measurements of the NEMA PET Body Phantom, as indicated by RCmean ≈ 1 for measurements independent of nuclide, contrast ratio, weight setup or sphere size.
RCmean and RCmax values of the smaller spheres were slightly lower in 68Ga-HCl compared to18F-FDG and 64Cu-HCl measurements, respectively, while parameters were comparable between 18F-FDG and 64Cu-HCl measurements. The maximal activity concentrations recovered from the PET/CT images were much higher than the true activity concentration in measurements of all nuclides, with lowest RCmax values in 68Ga-HCl measurements (1.22 ± 0.04) and highest RCmax in 64Cu-HCl measurements (1.51 ± 0.06). In 64Cu-HCl and 18F-FDG, but not in 68Ga-HCl measurements, the smallest sphere of 10 mm in diameter featured higher RCmax values than the remaining spheres.
For each nuclide, the percent contrast was comparable between different phantom setups and independent of the weight setup and sphere-to-background activity concentration ratio. The CRC was comparable between 18F-FDG and 64Cu-HCl measurements, but lower in 68Ga-HCl measurements (8:1 sphere-to-background activity concentration ratio and normal weight setup: 18F-FDG: 96.46% ± 4.07%; 64Cu-HCl: 103.90% ± 3.32%; 68Ga-HCl: 83.30% ± 5.75%) (Attachment, Table 1). At both contrast ratios and weight setups of the NEMA PET Body phantom, CNR was lower in 68Ga-HCl compared to 64Cu-HCl and 18F-FDG measurements, respectively.
Image noise was similar for all nuclides when comparing PET images of the NEMA PET Body phantom at the same setup (weight setup & sphere-to-background activity concentration ratio). Image noise was lower in normal weight setup compared to specific obese setup.
For both sphere-to-background activity concentrations, the relative count error in the lung insert was higher for 68Ga-HCl (4:1: 10.72%; 8:1: 12.62%) compared to 18F-FDG (4:1: 7.23%; 8:1: 7.75%) and 64Cu-HCl (8:1: 7.40%; 4:1: 7.29%), respectively.
Quantitative Evaluation of Impact of Reconstruction Method
Spatial resolution was best when standard reconstruction parameters were applied (4.10 mm ± 0.21 mm). Resolution was worse when ToF measurements were not performed (5.77 mm ± 0.59 mm) or when data was post-filtered with a Gaussian filter (3 mm FWHM: 4.89 mm ± 0.18 mm; 5 mm FWHM: 6.22 mm ± 0.18 mm). When PSF or ToF was not applied or data was post-filtered, RCmean, RCmax and CRC were not comparable between spheres of different sizes and especially the smaller spheres were not fully recovered (Figure 3). The influence of the ToF measurement was particularly noticeable in the recovery of the smallest sphere of 10 mm in diameter. Differences in RC, CRC and CNR values between the standard and modified reconstruction procedure were greatest when a Gaussian filter with 5 mm FWHM was applied. In comparison, a lower iteration number or a smaller matrix size did not affect image quality (Figure 3).
A reconstruction of the data without resolution recovery by PSF increased image noise substantially (OSEM+PSF: 9.7%; OSEM: 21.6%), while post-filtering of the data with a Gaussian filter reduced image noise (allpass: 9.7%; 3 mm FWHM: 6.6%; 5 mm FWHM: 4.2%). Similarly, CNRs were much lower in OSEM reconstruction (35.56 ± 2.67) compared to the OSEM + PSF reconstruction (86.55 ± 3.66) and much higher when post-filtering with a Gaussian filter and 5 mm FWHM was applied (155.98 ± 21.90) (Figure 3).