Phantom Study
Large area water ROI (see Fig.
6)
Tab.
1: HU values of the image part corresponding to water (mean ± SD)
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
3.5 ± 17.29 |
3.5 ± 14.77 |
3.5 ± 13.02 |
3.5 ± 10.72 |
CC |
3.05 ± 34.54 |
3.05 ± 28.63 |
3.05 ± 24.54 |
3.05 ± 19.37 |
CD |
8.8 ± 50.9 |
8.8 ± 42.23 |
8.8 ± 36.44 |
8.8 ± 29.27 |
Tab.
2: Minimal HU values of the image part corresponding to water
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
-94 |
-86 |
-82 |
-75 |
CC |
-212 |
-197 |
-191 |
-185 |
CD |
-289 |
-285 |
-276 |
-260 |
ROI in the water part (see Fig.
7)
Tab.
3: HU values of the ROI of the image part corresponding to water (mean ± SD)
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
1.70 ± 15.21 |
1.65 ± 12.62 |
1.66 ± 10.7 |
1.63 ± 8.25 |
CC |
1.86 ± 32.42 |
1.78 ± 26.6 |
1.76 ± 22.43 |
1.74 ± 17.2 |
CD |
4.89 ± 47.1 |
4.8 ± 38.6 |
4.75 ± 32.6 |
4.75 ± 25.14 |
Tab.
4: Minimal HU values of the ROI of the image part corresponding to water
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
-54 |
-45 |
-38 |
-31 |
CC |
-128 |
-107 |
-92 |
-72 |
CD |
-178 |
-147 |
-125 |
-98 |
Tab.
5: Maximal HU values of the ROI of the image part corresponding to water
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
55 |
48 |
43 |
37 |
CC |
117 |
100 |
87 |
72 |
CD |
170 |
140 |
120 |
95 |
ROI in the teflon part (see Fig.
8)
Tab.
6: HU values of the ROI in the teflon area (mean ± SD)
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
969.4 ± 22.4 |
968.4 ± 18.92 |
968.32 ±16.32 |
968 ± 13.79 |
CC |
968.9 ± 47.1 |
968.5 ± 39.3 |
968.86 ± 34.5 |
969.1 ± 28.42 |
CD |
1064.5 ± 70 |
1064.2 ± 58 |
1064.7 ± 51.7 |
1065.2 ± 43.3 |
Tab.
7: Minimal HU values of the ROI in the teflon area
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
874 |
883 |
889 |
895 |
CC |
778 |
797 |
808 |
825 |
CD |
794 |
834 |
849 |
874 |
Tab.
8: Maximal HU values of the ROI in the teflon area
convolution
kernel
|
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
CA |
1072 |
1061 |
1056 |
1046 |
CC |
1180 |
1157 |
1146 |
1127 |
CD |
1394 |
1354 |
1335 |
1314 |
As can be seen,
the mean values are preserved for different levels of the IR algorithm and FBP applying the same convolution kernel.
The standard deviations decrease with the level of IR as expected.
Applying different convolution kernels,
however,
leads to changes in the mean values.
Image quality assessment:
Tab.
9: Image quality parameters for convolution kernel (CA)
Parameter |
FBP |
IR level 2 |
IR level 4 |
IR level 6 |
MSE |
556 |
472 |
421 |
365 |
PSNR |
22.39 |
22.75 |
23 |
23.3 |
NAE |
0.02 |
0.017 |
0.0147 |
0.011 |
The mean square error (MSE) decreases with increasing level of iDose,
the peak signal to noise ratio (PSNR) slightly increases and the normalized absolute error (NAE) decreases as well.
The rather high error rates are caused by artifacts at the transition from teflon to water.
Myocardial perfusion indicators for segmented myocardial tissue
- Time-To-Peak: 14.8s for all reconstruction methods
- Wash-in-Rate varied from 9.74 to 9.94 HU/s
- Area-under-Curve (FBP/level3/level7): 1980/1972/1933 HUs
- Width (FBP/level3/level7): 17.77/17.76/17.78 s
Myocardial perfusion indicators are showing almost identical TTP,
Wash-in-Rate,
Area-Under-Curve and Width resulting in the same contrast enhancement time curve for all different reconstruction methods (compare alos Fig.
5).
Small ROIs in myocardial tissue
The small ROIs with only 64 voxels show similar contrast-enhancement-time curves,
with slightly lower noise in IR level 7.
In order to be able to demonstrate the small differences,
Fig.
8 shows the unfiltered HU-value time-curves for all three levels of IR.