A comprehensive study on the relationship between image quality...
2
A comprehensive study on the relationship between image quality and imaging dose in low-dose cone beam CT Introduction: CS-based CT reconstruction methods are able to handle both sparse-view reconstruction and low-mAs reconstruction, achieving a comparable image quality at a much lower imaging dose level. It is necessary to explore the relationship between image quality and imaging dose under this framework, which may provide guidance for designing optimal low-dose scan protocols for various specific IGRT tasks. Methods: 1) Data acquisition: Phantom was scanned with different tube load at 73.2-878.4 total mAs, Subsets of projections were extracted to get data sets (figure1a); Complementary simulation was performed to obtain data points covering the whole clinical range (figure 1b). 2) CBCT reconstruction: A GPU-implemented CS-based reconstruction algorithm with TF regularization was ran in a batch mode for efficient image reconstruction. 3) Image quality assessment (IQA): For the reconstructed volume images (figure 1c), IQA was performed by combining visual inspections and quantitative IQA metrics, including relative root mean square error (rRMSE 1 ), low-contrast emphasized RMSE (rRMSE 2 ), feature similarity index (FSIM), and contrast-noise ratio (CNR). 4) Dose-quality map (DQM) generation: The optimal IQA scores in different scan protocols were interpolated to get the so called DQMs. a) b) c) Figure 1. a) The measurement data grid points and four iso-dose lines. Only the dense grid points inside the red box will be used due to accuracy considerations; b) The simulation data grid points and six iso-dose lines; c) Modules and region of interests (ROIs) in the phantom that were investigated, including a contrast slice, the volume containing low-contrast and high contrast objects, seven ROIs in the contrast slice, the high resolution slice and the zoomed ROI. Results: 1) DQM and image quality on iso-dose lines DQMs from simulated and experimental data in terms of rRMSE 1 , rRMSE 2 , and FSIM are shown in figures 2 and 3 respectively, in which x-axis is projection number and y-axis is imaging dose, and the color represents the IQA score values. From figure 2, we have following observations: 1) Image quality has little degradation over a large imaging dose range (upper-right corner) but degrades rapidly at the low dose range below 100 total mAs (lower-left corner), and degrades dramatically when the dose is below 40 total mAs (pink dashed line, figure 2b). 2) For cases with extremely sparse views (projection number < 50), image quality degrades severely no matter how high the mAs/view is. 3) There exists a point of optimal image quality on an iso-dose line. This point is near the intersection of each iso-dose line and steepest image quality variation direction (white dot line in the 3 rd image of figure 2a), indicating that the optimal scan protocol in low- dose range consists of a medium number of projections and a medium mAs/view level. These observations especially observation 3) were further verified by the experimental data, such as the consistent DQMs (figure 3a), the global IQA scores along iso-dose lines (figure 3b) and CNR scores for each ROI (figure 3c). Figure 3c shows that the observation 3) is more obvious for low contrast objects at lower dose levels (yellow and red) while not obvious for the high-contrast objects (blue and cyan). CNR decreases only for the low- contrast ROIs (black, yellow, and red) and does not change much for high-contrast ROIs. 46 61 91 121 182 364 0.2 0.4 0.6 0.9 1.2 1.6 2.4 Proj Number mAs / View 36.4mAs 72.8mAs 109.2mAs 145.6mAs 30 90 150 210 270 330 360 0.2 0.6 1.2 1.6 Proj Number mAs level ROI#3 ROI#5 ROI#4 ROI#6 ROI#7 ROI#1 ROI#2
Transcript of A comprehensive study on the relationship between image quality...
A comprehensive study on the relationship between image quality and imaging dose in low-dose cone beam CT
Introduction: CS-based CT reconstruction methods are able to handle both sparse-view reconstruction and low-mAs reconstruction, achieving a comparable image quality at a much lower imaging dose level. It is necessary to explore the relationship between image quality and imaging dose under this framework, which may provide guidance for designing optimal low-dose scan protocols for various specific IGRT tasks. Methods: 1) Data acquisition: Phantom was scanned with different tube load at 73.2-878.4 total mAs, Subsets of projections were extracted to get data sets (figure1a); Complementary simulation was performed to obtain data points covering the whole clinical range (figure 1b). 2) CBCT reconstruction: A GPU-implemented CS-based reconstruction algorithm with TF regularization was ran in a batch mode for efficient image reconstruction. 3) Image quality assessment (IQA): For the reconstructed volume images (figure 1c), IQA was performed by combining visual inspections and quantitative IQA metrics, including relative root mean square error (rRMSE1), low-contrast emphasized RMSE (rRMSE2), feature similarity index (FSIM), and contrast-noise ratio (CNR). 4) Dose-quality map (DQM) generation: The optimal IQA scores in different scan protocols were interpolated to get the so called DQMs.
a) b)
c)
Figure 1. a) The measurement data grid points and four iso-dose lines. Only the dense grid points inside the red box will be used due to accuracy considerations; b) The simulation data grid points and six iso-dose lines; c) Modules and region of interests (ROIs) in the phantom that were investigated, including a contrast slice, the volume containing low-contrast and high contrast objects, seven ROIs in the contrast slice, the high resolution slice and the zoomed ROI.
Results: 1) DQM and image quality on iso-dose lines DQMs from simulated and experimental data in terms of rRMSE1, rRMSE2, and FSIM are shown in figures 2 and 3 respectively, in which x-axis is projection number and y-axis is imaging dose, and the color represents the IQA score values. From figure 2, we have following observations: 1) Image quality has little degradation over a large imaging dose range (upper-right corner)
but degrades rapidly at the low dose range below 100 total mAs (lower-left corner), and degrades dramatically when the dose is below 40 total mAs (pink dashed line, figure 2b).
2) For cases with extremely sparse views (projection number < 50), image quality degrades severely no matter how high the mAs/view is.
3) There exists a point of optimal image quality on an iso-dose line. This point is near the intersection of each iso-dose line and steepest image quality variation direction (white dot line in the 3rd image of figure 2a), indicating that the optimal scan protocol in low-dose range consists of a medium number of projections and a medium mAs/view level.
These observations especially observation 3) were further verified by the experimental data, such as the consistent DQMs (figure 3a), the global IQA scores along iso-dose lines (figure 3b) and CNR scores for each ROI (figure 3c). Figure 3c shows that the observation 3) is more obvious for low contrast objects at lower dose levels (yellow and red) while not obvious for the high-contrast objects (blue and cyan). CNR decreases only for the low-contrast ROIs (black, yellow, and red) and does not change much for high-contrast ROIs.
4661 91 121 182 3640.20.40.6
0.9
1.2
1.6
2.4
Proj Number
mA
s / V
iew
36.4mAs72.8mAs109.2mAs145.6mAs
30 90 150 210 270 330 3600.2
0.6
1.2
1.6
Proj Number
mA
s le
vel
ROI#3
ROI#5ROI#4
ROI#6
ROI#7ROI#1
ROI#2
a)
b) Figure 2 Results of simulated data. a). DQMs in terms of rRMSE1, rRMSE2, and FSIM from left to right, with six iso-dose lines at total mAs levels from 24.3 to 145.6. The white dot line with arrow head in the rightmost image indicates the steepest image quality variation direction. b) The variation of the IQA scores with the total mAs, along the white dot line in the 3rd image of a). The pink dashed vertical line indicates 40 total mAs below which the IQA scores degrade at a steeper rate.
a)
b) c)
Figure 3 Results of experimental data. a) DQMs in terms of rRMSE1, rRMSE2, and FSIM from left to right, with four iso-dose lines around 36.4, 72.8, 109.2, and 145.6 total mAs superimposed. b) IQA scores for the images containing low-contrast
objects at the above four iso-dose levels. c). CNR scores for ROIs with different contrast at four iso dose levels
2) Extremely low dose level for different IGRT tasks The reconstructed images acquired at extremely low dose levels are shown in figure 4, by referring images acquired at 873.6 total mAs. It shows that neither 12.2 nor 18.2 total mAs is sufficient for the low-contrast or high-resolution objects. However, with even 12.2 total mAs we can still clearly see the high-contrast objects with a dimension of. e.g., 3 mm diameters (small Teflon and air rob). This is consistent with the CNR results for the high-contrast objects shown in figure 3c), i.e., the visibility of high-contrast ROIs suffers the least from the dose reduction. Therefore, it may be quite feasible to use an extremely low dose level like 12.2 total mAs for certain IGRT tasks such as bony structure based patient positioning.
high-contrast low-contrast high-resolution 873.6 18.2 12.2 873.6 18.2 12.2 873.6 18.2 12.2
Figure 4. Left column: images with high-contrast objects with display grayscale [0.008, 0.032] mm-1; middle column: images with low-contrast objects (as indicated by the yellow arrows) with display grayscale [0.018, 0.025] mm-1; right column: high-resolution ROIs with display grayscale [0.03, 0.05] mm-1; In each column, left: highest dose level used in the experiment (873.6 total mAs); middle and right : extremely low dose levels (18.2 and 12.2 total mAs, respectively).
50 100 150 200 250 300 3500.2
0.4
0.6
0.8
1
1.2
1.4
Proj Number
mA
s / V
iew
0.005
0.01
0.015
0.02
0.02524.3mAs36.4mAs54.6mAs72.8mAs109.2mAs145.6mAs
50 100 150 200 250 300 3500.2
0.4
0.6
0.8
1
1.2
1.4
Proj Number
mA
s / V
iew
0.02
0.04
0.06
0.08
0.1
0.12
0.1424.3mAs36.4mAs54.6mAs72.8mAs109.2mAs145.6mAs
50 100 150 200 250 300 3500.2
0.4
0.6
0.8
1
1.2
1.4
Proj Number
mA
s / V
iew
0.98
0.985
0.99
0.995
24.3mAs36.4mAs54.6mAs72.8mAs109.2mAs145.6mAs
50 100 150 200 250 3000.975
0.980
0.985
0.990
0.995
1.000
FSIM
0
0.050
0.100
0.150
0.200
0.250
Total mAs
rRSM
E
FSIM10rRMSE
1
rRMSE2
50 100 1500.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Proj Number
mA
s / V
iew
0
0.005
0.01
0.015
0.02
36.4mAs72.8mAs109.2mAs145.6mAs
50 100 1500.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Proj Number
mA
s / V
iew
0
0.1
0.2
0.3
0.4
0.5
0.636.4mAs72.8mAs109.2mAs145.6mAs
50 100 1500.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Proj Number
mA
s / V
iew
0.97
0.975
0.98
0.985
0.99
0.995
136.4mAs72.8mAs109.2mAs145.6mAs
0 200 4000
24
0.005
0.01
0.015
0.02
Proj NumbermAs / View
RM
SE
1
0200
400
02
40
0.1
0.2
0.3
0.4
Proj NumbermAs / View
RM
SE
2
0 200 400024
0.985
0.99
0.995
1
Proj NumbermAs / View
FS
IM
36.4mAs72.8mAs109.2mAs145.6mAs
0 200 400
2
4
6
Proj Number
CN
R
36.4mAs
roi#1roi#2roi#3roi#4roi#5roi#6roi#7
0 200 400
2
4
672.8mAs
0 200 400
2
4
6109.2mAs
0 200 400
2
4
6145.6mAs
363.8 364 364.2
2
4
6873.6mAs
