Special Ar ticles • Original Research Nitta et al. Quality of 320-MDCT Images of Abdominal Phantom
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Special Articles Original Research
Norihisa Nitta1 Mitsuru Ikeda2 Akinaga Sonoda1 Yukihiro Nagatani1 Shinichi Ohta1 Masashi Takahashi1 Kiyoshi Murata1 Nitta N, Ikeda M, Sonoda A, et al.
Images Acquired Using 320-MDCT With Adaptive Iterative Dose Reduction With Wide-Volume Acquisition: Visual Evaluation of Image Quality by 10 Radiologists Using an Abdominal Phantom OBJECTIVE. The purpose of this study is to assess visual evaluations of CT images and to determine by how much radiation exposure dose could be reduced without compromising the image quality. MATERIALS AND METHODS. An abdominal CT phantom was scanned at 14 different tube currents. Raw data were reconstructed with adaptive iterative dose reduction (AIDR) 3D and filtered backprojection (FBP). We divided 64 different image pairs into five groups. Group A consisted of 14 image pairs acquired with AIDR 3D and FBP, groups B and D consisted of 13 pairs with a one-level exposure dose decrease in AIDR 3D and FBP, respectively, and groups C and E consisted of 12 pairs with a two-level exposure dose decrease in AIDR 3D and FBP, respectively. Ten radiologists participated in the reading session. Statistical analyses were calculated with analysis of variance and the paired Student t test. RESULTS. Analysis of variance of six criteria revealed that the results were better in groups A, D, and E when AIDR 3D was applied. Better results were obtained with FBP in groups B and C. When we subjected evaluations of the renal parenchyma to the Student t test, we found that the assigned scores were better with AIDR 3D in groups A, D, and E and better with FBP in groups B and C. Similar results were obtained for the other evaluation criteria. CONCLUSION. Visual subjective evaluation showed that images of acceptable quality could be obtained at dose reductions of approximately 10% in the high-dose range and about 20% in the moderate-dose range.
A
Keywords: 320-MDCT, abdominal phantom, filtered backprojection, image quality, iterative dose reduction DOI:10.2214/AJR.12.10364 Received December 1, 2012; accepted after revision April 3, 2013. 1 Department of Radiology, Shiga University of Medical Science, Tsukinowa-cho Seta, Otsu Shiga 520-2192, Japan. Address correspondence to N. Nitta. 2 Division of Radiological Technology, School of Health Sciences Radiological Technology, Nagoya University, Nagoya Aichi, Japan.
AJR 2014; 202:2–12 0361–803X/14/2021–2 © American Roentgen Ray Society
2
lthough MDCT studies yield clinically useful images, the radiation exposure dose tends to increase with diagnostic accuracy [1–4]. Because radiation exposure at CT may increase the risk for cancer in populations of a specific age and sex [1], it must be reduced without affecting image quality. To this end, manufacturers of CT scanners have developed iterative reconstruction techniques [5, 6], such as AIDR (adaptive iterative dose reduction; Toshiba Medical Systems), ASiR (adaptive statistical iterative reconstruction; GE Healthcare), iDose4 (Philips Healthcare), and SAFIRE (sinogram-affirmed iterative reconstruction; Siemens Healthcare). Although exploratory studies have shown that iterative reconstruction techniques facilitate a 20–30% dose reduction without compromising image quality [7–9], detailed data are lacking. Therefore, we used paired upper abdominal phantom images to compare the quality of iterative reconstruction CT images obtained at various radiation exposure doses
and of CT images acquired with filtered backprojection (FBP) at the same, higher, and lower doses. Materials and Methods Abdominal Phantom We used a commercially available upper abdominal CT phantom (PH-5, Kyoto Kagaku) consisting of bone and soft tissue (Fig. 1). The bone tissue is composed of epoxy resin, calcium carbonate (for the part corresponding to the medullary bone), and barium sulfate (for cortical bone). The soft tissue is made of hydroxyapatite and polyurethane resin whose mixture rate is adjusted to yield the same CT number (in Hounsfield units) as the corresponding abdominal organ. The longest and shortest diameter of the phantom are 26 and 18 cm, respectively, and the abdominal girth is 79 cm. The CT numbers of images of this phantom are 400 HU for vertebrae and ribs; 70 HU for the liver parenchyma; 50 HU for the spleen; 40 HU for portal and hepatic veins, the aorta, inferior vena cava, and other blood vessels; 30 HU for the pancreas and kidneys; 20 HU for the gallbladder; and −7 HU for soft tissue around organs.
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
CT Image Acquisition Using an abdominal CT phantom and a 320MDCT scanner (Aquillon ONE, Toshiba Medical Systems), we acquired images reconstructed with AIDR 3D (a modified AIDR technique developed by Toshiba Medical Systems) and with conventional FBP [10]. AIDR is a technology to selectively extract and remove only noise other than signals on the 3D voxels of reconstructed images [11]. AIDR 3D efficiently removes noise and artifacts from raw data using a statistical noise and scanner model. It extracts only noise and repeatedly removes noise at image reconstruction with an anatomic model [12]. We applied 14 levels of tube current, ranging from 20 to 550 mA. Other scanning parameters were tube voltage of 120 kV, collimation of 0.5 mm, scan time of 0.5 seconds, reconstruction kernel of FC04, and slice thickness of 5 mm. Automatic exposure control was not used in this study. In one rotation, we acquired CT images of the abdominal phantom (width, 16 cm) from the liver beneath the diaphragm to the lower pole of the kidney. We scanned each slice once at each scanning condition to obtain image pairs; one image was reconstructed with AIDR 3D and the other with the conventional FBP method. The reconstructed images were transferred to an imageviewing workstation (Ziostation, Ziosoft). The radiation absorption dose (exposure dose) was evaluated by CT dose index displayed on the console of the CT system. At our institute, the CT dose index is set at 20 mGy for routine abdominal CT examinations [13]. For convenience, we considered CT dose index at, above, and below approximately 20 mGy as moderate-, high-, and lowexposure radiation dose ranges, respectively.
Noise Evaluation To measure the image noise, we placed a circular 4-cm 2 region of interest (ROI) on five sites on each of 14 phantom CT images of the liver. ROIs were placed so that they were evaluated at exactly the same site in all 14 CT series by using a copyand-paste feature. These were acquired at varying tube currents, and the SD of the CT number in each ROI was calculated. We refer to this value as the noise evaluation.
Image Evaluation We prepared 64 image pairs; one CT image in each pair was reconstructed with AIDR 3D, and the corresponding image was reconstructed with FBP. Then we divided the image pairs into five groups as shown in Table 1. Group A contained 14 basic image pairs that were acquired at the same radiation exposure dose, ranging from 1.0 to 29.2 mGy. In groups B (13 pairs) and C (12 pairs), the
radiation exposure for the AIDR 3D–reconstructed image was one and two levels lower, respectively, than for the FBP image. In groups D (13 pairs) and E (12 pairs), the radiation exposure for the AIDR 3D–reconstructed image was one and two levels higher, respectively, than for the FBP image. The image pairs were displayed in landscape orientation on two high-resolution 24-inch monitors (MultiSync LCD2490WUXi2, NEC) placed side-by-side. The display conditions on the monitors were identical. The 64 image pairs were interpreted by 10 board-certified radiologists with 9–18 years of experience; they were blinded to the reconstruction methods and acquisition parameters. All readers understood the purpose of the study. The readers independently performed direct sideby-side comparisons of each CT image pair. Differences with respect to our seven evaluation criteria (visibility of intrahepatic vessels, discrimination between pancreas and splenic vein, visibility of the pancreatic margin, visibility of the margin between the inferior vena cava and aorta, strength of streak artifact from the vertebra, visibility of the renal parenchyma, and overall judgment) were recorded in pencil on a continuous 20-cm-long line corresponding to a Likert scale ranging from −5 to 5 (left and right end of the line, respectively), where −5 to −3 means that the left image is better, −3 to −1 means that the left image is slightly better, −1 to 1 means that the images are almost equivalent, 1 to 3 means that the right image is slightly better, and 3 to 5 means that the right image is better. Because the same image pair was evaluated twice with the display side changed, each observer read a total of 128 image pairs. These were displayed in random order; no two image pairs were displayed successively in the reading order. At each side-by-side comparison, the absolute value of the rating scale was assigned to the better image with a value of 0 assigned to the image on the other side. For image interpretation, the window and level were fixed at 350 and 40 HU, respectively. Viewing time was unlimited.
Statistical Analysis Excluding overall judgment, there were six evaluation criteria. These were defined as criteria A. To compare the evaluation scores of image pairs obtained with the two reconstruction methods, we performed analysis of variance (ANOVA) for a two repeated-measures design, using the obtained scores as the dependent variable, the two reconstruction methods as the within-subjects factor, and criteria A as the between-subjects factor [14]. If the null hypothesis in the Mauchly test of sphericity was rejected, we adjusted the df of F test statistics using the Greenhouse-Geisser test, the Huynh-Feldt test, and the lower-bound epsilon. For the statistical tests,
we adopted a significance level of 0.05. For this series of ANOVA tests, the adopted level of significance was 0.00078125 (0.05/64) to compensate for the multiple comparisons problem by the Bonferroni correction. For the interaction effect test, we did not use the Bonferroni correction; rather, we adopted a level of significance of 0.05 to decrease type 2 errors and to increase the statistical power. We also performed paired Student t tests to compare the evaluation scores for each of the six evaluation criteria plus overall judgment (criteria B) that were assigned to the image pairs reconstructed with the two different methods. To compensate for the multiple comparisons problem by the Bonferroni correction, we adopted a level of significance of 0.000111607 (0.05/448) for this series of paired Student t tests. To evaluate differences in the evaluation scores of images displayed on the right and left sides, we subjected all images, each image pair, and each criterion of criteria B to the paired Student t test. For this series of paired Student t tests, we did not use the Bonferroni correction and adopted a level of significance of 0.05 to decrease type 2 errors and to increase the statistical power. SSPS software (version 17, SPSS) was used for all statistical analyses.
Results On all 14 image pairs obtained at a dose ranging from 1.0 to 29.2 mGy, the noise evaluation of the AIDR 3D–reconstructed image was lower than that of the corresponding FBPreconstructed image. The maximum noise reduction was approximately 65% (Table 2). Table 3 shows the mean quality scores assigned for criteria A to each of the image pairs in the five groups. In all group A pairs, the quality of the AIDR 3D image was rated as better than that of the corresponding FBP image. The difference was statistically significant in four low-dose and two moderatedose pairs. Because, with the exception of one image pair, ANOVA revealed no statistically significant interaction between the reconstruction method and the evaluation criterion, this statistical difference was almost independent of the evaluation criterion except for the one image pair. On the other hand, in nine group B and 11 group C pairs, the quality of the AIDR 3D image was lower than that of the FBP image; this difference was statistically significant in three low-dose pairs of group B and all pairs of group C (Table 4). There was no statistically significant interaction between the reconstruction method and the evaluation criterion. In 11 group D and all 12 group E pairs, the quality of the AIDR 3D image was higher than
AJR:202, January 2014 3
Nitta et al. TABLE 1: CT Dose Index (CTDI), Tube Current, and Radiation Dose Reduction Rate of 64 Image Pairs AIDR-Reconstructed CT Image Group, Image Pair
CTDI (mGy)
Tube Current (mA)
FBP-Reconstructed CT Image CTDI (mGy)
Tube Current (mA)
Dose Reduction Rate (%)
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
29.2
550
29.2
550
0.0
2
26.6
500
26.6
500
0.0
3
24.4
460
24.4
460
0.0
4
22.3
420
22.3
420
0.0
5
20.2
380
20.2
380
0.0
6
18.1
340
18.1
340
0.0
7
15.9
300
15.9
300
0.0
8
13.1
260
13.1
260
0.0 0.0
9
11.1
220
11.1
220
10
9.1
180
9.1
180
0.0
11
7.0
140
7.0
140
0.0
12
4.5
90
4.5
90
0.0
13
2.5
50
2.5
50
0.0
14
1.0
20
1.0
20
0.0
Group B 1
26.6
500
29.2
550
8.9
2
24.4
460
26.6
500
8.3 8.6
3
22.3
420
24.4
460
4
20.2
380
22.3
420
9.4
5
18.1
340
20.2
380
10.4
6
15.9
300
18.1
340
12.2
7
13.1
260
15.9
300
17.6
8
11.1
220
13.1
260
15.3
9
9.1
180
11.1
220
18.0
10
7.0
140
9.1
180
19.1
11
4.5
90
7.0
140
35.7
12
2.5
50
4.5
90
44.4
13
1.0
20
2.5
50
60.0
1
24.4
460
29.2
550
16.4
2
22.3
420
26.6
500
16.2
3
20.2
380
24.4
460
17.2
4
18.1
340
22.3
420
18.9
Group C
5
15.9
300
20.2
380
21.3
6
13.1
260
18.1
340
27.6
7
11.1
220
15.9
300
30.2
8
9.1
180
13.1
260
30.5
9
7.0
140
11.1
220
36.9
10
4.5
90
9.1
180
50.1
11
2.5
50
7.0
140
64.3
12
1.0
20
4.5
90
77.8
Note—The reconstruction methods used in groups B and C were reversed in groups D and E. AIDR = adaptive iterative dose reduction, FBP = filtered backprojection.
4
AJR:202, January 2014
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Quality of 320-MDCT Images of Abdominal Phantom that of the FBP image. This difference was statistically significant in the eight group D and 11 group E image pairs. There was no statistically significant interaction between the reconstruction method and the evaluation criterion. For all evaluation criteria except overall judgment (criteria B), the assigned scores were similar to the mean scores. Figure 2 is an illustrative example. It shows the score assigned for visibility of the renal parenchyma to each of the 64 image pairs. Table 4 shows the mean quality scores assigned for the renal parenchyma to each of the image pairs in the five groups. With the exception of the highest-dose image pair, in 13 group A pairs, visibility of the renal parenchyma was scored as better on AIDR 3D images than on FBP images. This difference was statistically significant in the two low-dose pairs. On the other hand, in five group B and all group C pairs, visibility of the renal parenchyma was judged to be better on FBP images than on AIDR 3D images. This difference was statistically significant in two and three low-dose pairs of groups B and C, respectively. For all image pairs of groups D and E, visibility of the renal parenchyma was scored to be better on AIDR 3D images than on FBP images. The difference was statistically significant in four low-dose pairs of group D and one moderate-dose and five low-dose pairs of group E. Similar results were obtained for the other evaluation criteria. In fact, the quality of the image displayed on the left side was judged to be significantly better than that on the right side (p < 0.000001) (Fig. 3). There was a statistically significant difference in the quality score assigned to images displayed on the left and right sides in 55 of the total 896 evaluations (64 images × 2 sides × 7 criteria = 896 evaluations), although we detected no particular trend to explain this observation. In some cases, the image quality scores assigned to each side were identical. Each image pair was evaluated twice with the display side reversed to compensate for bias. Because our search for potential factors contributing to this bias detected none, we think that the bias is incidental. Discussion Our study shows that, by the subjective judgment of 10 readers, the quality of CT images reconstructed with AIDR 3D was better than that of FBP-reconstructed images obtained at the same radiation exposure dose. This difference in quality was more evident at lower radiation doses. On the other hand,
TABLE 2: Noise Evaluation of 14 Phantom CT Images Image No.
Radiation Exposure Dose (mGy)
1 2
Image Noise (Mean ± SD), Reconstruction Method
Tube Current (mA)
AIDR
FBP
29.2
550
5.80 ± 0.48
8.70 ± 0.58
26.6
500
6.22 ± 0.40
8.62 ± 1.12
3
24.4
460
6.56 ± 0.53
9.62 ± 0.66
4
22.3
420
6.68 ± 0.55
9.78 ± 0.50 10.36 ± 0.62
5
20.2
380
7.00 ± 0.52
6
18.1
340
7.52 ± 0.89
11.08 ± 1.01
7
15.9
300
8.10 ± 0.64
12.04 ± 0.79
8
13.1
260
8.44 ± 0.85
12.82 ± 1.17
9
11.1
220
8.70 ± 0.52
13.58 ± 0.82
10
9.1
180
9.68 ± 0.74
15.42 ± 1.23
11
7.0
140
10.52 ± 0.74
17.38 ± 1.74
12
4.5
90
12.48 ± 0.70
22.36 ± 1.74
13
2.5
50
17.22 ± 1.80
33.64 ± 3.64
14
1.0
20
25.56 ± 1.68
63.78 ± 6.87
Note—Noise evaluation was calculated from the SD of the CT numbers in each 4-cm2 circular region of interest placed at five sites on the liver in each phantom CT image. AIDR = adaptive iterative dose reduction, FBP = filtered backprojection.
there were no instances in which the subjectively judged quality of AIDR 3D images was significantly better than that of FBP images obtained at the higher dose, and there were instances in which the quality of AIDR 3D images was judged to be statistically worse than that of FBP images obtained at the higher radiation dose. Consequently, we cannot adduce evidence for the contention that the AIDR 3D method allows a reduction in the radiation dose without a detrimental effect on the image quality. If the probability of the null hypothesis for a statistical test is posited to be above 0.1, then the quality of image pairs is almost equivalent for the statistical test. Under this hypothesis, our ANOVA results for criteria A suggest that the subjectively judged quality of members of image pairs 1, 2, 3, 5, and 7 of group B and of the members of pair 1 in group C is almost equivalent (Table 3). In other words, with AIDR 3D, a dose reduction of 9–17% can be obtained without rendering the subjective quality of images worse than that of images acquired at high and standard radiation doses. The results of ANOVA for criteria A also indicate that the subjective image quality can be considered to be almost equivalent in image pairs 1, 2, and 5 of group D in Table 3. That is, the quality of AIDR 3D CT images acquired at the high and moderate radiation
dose will also be almost equivalent to that of FBP images obtained with a dose reduction of 9–10%. On the basis of these considerations, we suggest that the subjective image quality evaluation of CT images obtained at the high dose level is robust. According to the noise evaluation, the noise level of FBP-reconstructed CT images scanned at 29.2, 22.3, 13.1, and 7.0 mGy was almost equivalent to that of AIDR 3D–reconstructed images scanned at 11.1, 9.1, 4.5, and 2.5 mGy, respectively. This finding suggests that, compared with FBP, AIDR 3D reduces the radiation exposure dose by 60–65% at image acquisition based solely on the noise level. However, contrary to our expectations, we found that, in almost all instances, the quality of AIDR 3D images obtained with a dose reduction above 25% was subjectively judged to be worse than that of FBP images. This difference was statistically significant in all cases where the dose reduction exceeded 60%. Thus, a dose reduction of 9–17% may be feasible at upper abdominal CT performed at the standard- and high-dose level when the images are reconstructed with AIDR 3D. In studies on tolerable dose reduction rates for iterative reconstruction techniques, the reduction rate was higher [15, 16]. Although two or three readers participated in most other studies on the subjective quality of CT images reconstructed with iterative reconstruction techniques [15,
AJR:202, January 2014 5
Nitta et al. TABLE 3: Image Quality Evaluation Scores on Six Evaluation Criteria Excluding Overall Judgment (Criteria A) Scores for AIDR-Reconstructed Images Group, Image
Mean
Scores for FBP-Reconstructed Images
SD
Mean
p
SD
ANOVA
Interaction
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
0.48
0.73
0.38
0.75
0.40
0.91
2
0.56
0.68
0.27
0.50
0.0018
0.33
3
0.49
0.73
0.26
0.53
0.02
0.94
4
0.48
0.73
0.31
0.58
0.10
0.80
5
0.50
0.65
0.28
0.65
0.03
0.71
6
0.57
0.77
0.21
0.47
0.0002
0.98
7
0.47
0.57
0.16
0.37
0.00003
0.96
8
0.41
0.58
0.26
0.47
0.06
0.78 0.92
9
0.65
0.83
0.25
0.80
0.001
10
0.61
0.84
0.26
0.58
0.002
0.78
11
0.61
0.59
0.11
0.29
1.50 × 10 −10
0.89
12
0.73
0.92
0.22
0.48
1.18 × 10 −5
0.86
13
0.88
0.73
0.02
0.13
0.00
0.03
14
1.20
1.06
0.004
0.04
0.00
0.23
Group B 1
0.35
0.55
0.30
0.48
0.51
0.56
2
0.33
0.59
0.28
0.40
0.49
0.14
3
0.31
0.47
0.37
0.67
0.50
0.95
4
0.15
0.27
0.38
0.62
0.001
0.55
5
0.32
0.52
0.32
0.51
0.94
0.91
6
0.23
0.46
0.32
0.48
0.19
0.32
7
0.38
0.57
0.35
0.59
0.74
0.76
8
0.24
0.43
0.45
0.58
0.009
0.09
9
0.26
0.45
0.46
0.54
0.02
0.96
10
0.35
0.50
0.29
0.45
0.42
0.53
11
0.11
0.33
1.00
0.89
6.6 × 10 −16
0.17
12
0.02
0.11
1.27
1.14
0.00
0.61
13
0.01
0.06
2.52
1.59
0.00
0.61
1
0.30
0.58
0.22
0.38
0.31
0.67
2
0.09
0.20
0.43
0.54
4.27 × 10 −8
0.15 0.19
Group C
3
0.08
0.33
0.79
0.90
5.06 × 10 −12
4
0.17
0.38
0.42
0.52
4.00 × 10 −4
0.45
5
0.12
0.31
0.45
0.49
2.70 × 10 −7
0.07
6
0.12
0.24
0.45
0.57
8.69 × 10 −7
0.19 0.14
7
0.07
0.19
0.33
0.43
7.55 × 10 −8
8
0.08
0.24
0.70
0.76
1.08 × 10 −12
0.08
9
0.22
0.72
0.68
0.80
5.87 × 10 −5
0.65
10
0.004
0.05
1.56
1.18
0.00
0.75
11
0.02
0.13
2.45
1.36
0.00
0.86
12
0.00
0.00
3.34
1.41
0.00
0.94 (Table 3 continues on next page)
6
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom TABLE 3: Image Quality Evaluation Scores on Six Evaluation Criteria Excluding Overall Judgment (Criteria A) (continued) Scores for AIDR-Reconstructed Images Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group, Image
Scores for FBP-Reconstructed Images
p
Mean
SD
Mean
SD
ANOVA
Interaction
0.29
0.55
0.39
0.65
0.26
0.66
Group D 1 2
0.42
0.59
0.32
0.59
0.26
0.26
3
0.56
0.74
0.33
0.56
0.03
0.99
4
0.43
0.63
0.13
0.26
4.53 × 10 −5
0.93
5
0.44
0.65
0.38
0.69
0.58
0.73
6
0.36
0.59
0.46
0.71
0.35
0.86
7
0.60
0.71
0.26
0.57
7.40 × 10 −4
0.82 0.79 0.24
8
0.89
0.81
0.12
0.38
8.82 × 10 −13
9
0.94
0.87
0.06
0.20
0.00
10
0.84
0.99
0.25
0.59
5.35 × 10 −6
0.98
11
2.07
1.09
0.002
0.02
0.00
0.90
12
2.52
1.29
0.04
0.45
0.00
0.26
13
3.70
1.23
0.00
0.00
0.00
0.95
1
0.43
0.61
0.21
0.42
8.00 × 10 −3
0.78
2
0.45
0.62
0.17
0.30
3.00 × 10 −4
0.66
3
0.51
0.67
0.21
0.40
4.00 × 10 −4
0.22
4
0.56
0.72
0.19
0.45
6.41 × 10 −5
0.99
Group E
5
0.83
0.65
0.03
0.15
0.00
0.57
6
0.88
0.86
0.08
0.22
3.11 × 10 −15
0.25
7
0.88
0.87
0.09
0.25
5.45 × 10 −14
0.73
8
1.67
1.11
0.03
0.10
0.00
0.93
9
1.39
0.90
0.02
0.14
0.00
0.49
10
2.55
1.25
0.003
0.04
0.00
0.67
11
3.30
1.24
0.00
0.00
0.00
0.82
12
4.08
0.94
0.02
0.18
0.00
0.84
Note—AIDR = adaptive iterative dose reduction, ANOVA = analysis of variance, FBP = filtered backprojection.
17–21], 10 experienced radiologists evaluated our phantom CT images independently. Consequently, the power of our statistical tests is higher than that of the other studies. In addition, we performed direct comparisons by displaying image pairs side by side. We think that the dose reduction rate that is possible with iterative reconstruction techniques without affecting the image quality is lower than has been reported in the above-cited studies. Because our readers performed subjective evaluations of the quality of images displayed side by side, we cannot rule out the possibility of display-side bias. In fact, images displayed on the left side received significantly better evaluation scores. To compensate for this bias, each reader evaluated each
image pair twice with the display side of the images reversed. Our study has some limitations. First, training data were not provided to the radiologists at the image reading sessions and only the image evaluation method was explained. Second, only unenhanced phantom CT images were evaluated, and clinical and contrast-enhanced CT images must be included in future evaluation studies. Third, the range of the radiation doses we were able to use was limited and was determined by our CT system. Fourth, as in other studies on the quality of CT images reconstructed with iterative reconstruction techniques [8, 22, 23], we did not investigate the image quality necessary for a clinical diagnosis. We ex-
pect that the radiation dose reduction without a significant effect on the diagnosis will be greater than the dose reduction that does not affect the subjective evaluation of image quality [24–27]. Conclusion At the same radiation exposure dose, the subjectively judged quality of phantom upper abdominal CT images reconstructed with AIDR 3D was higher than that of images reconstructed with FBP. By use of AIDR 3D reconstruction, the dose for upper abdominal CT scans can be reduced by 9–17% of the standard- and high-dose level without degradation of the subjective image quality.
AJR:202, January 2014 7
Nitta et al. TABLE 4: Image Quality Evaluation Score on Visibility of Renal Parenchyma Scores for AIDR-Reconstructed Images Group, Image No.
Mean
SD
Scores for FBP-Reconstructed Images Mean
SD
p for ANOVA
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
0.42
0.60
0.50
0.99
0.79
2
0.69
0.82
0.21
0.45
0.06
3
0.58
0.72
0.25
0.51
0.18
4
0.55
0.90
0.31
0.50
0.38
5
0.54
0.66
0.11
0.21
0.02
6
0.66
0.80
0.14
0.38
0.03
7
0.52
0.63
0.19
0.42
0.11
8
0.53
0.69
0.15
0.30
0.06
9
0.70
0.83
0.23
0.40
0.07
10
0.85
0.88
0.21
0.51
0.03
11
0.70
0.56
0.09
0.27
1.00 × 10 −3
12
1.02
1.13
0.22
0.46
13
1.22
0.89
0
0
7.49 × 10 −6
0.02
14
1.39
1.26
0
0
9.25 × 10 −5
Group B 1
0.44
0.57
0.19
0.30
0.15
2
0.37
0.53
0.30
0.45
0.73
3
0.28
0.47
0.31
0.64
0.88
4
0.16
0.26
0.30
0.44
0.33
5
0.37
0.55
0.37
0.52
1.00
6
0.32
0.50
0.17
0.37
0.36
7
0.44
0.54
0.21
0.42
0.22
8
0.29
0.40
0.22
0.34
0.65
9
0.32
0.51
0.45
0.54
0.55
10
0.40
0.58
0.36
0.48
0.87
11
0.22
0.46
0.86
0.70
0.01
12
0
0
1.14
0.97
4.70 × 10 −5
13
0
0
2.72
1.52
1.67 × 10 −7
Group C 1
0.24
0.48
0.28
0.41
0.81
2
0.06
0.17
0.56
0.59
3.00 × 10 −3
3
0.11
0.42
0.65
0.81
0.02
4
0.22
0.43
0.39
0.41
0.31
5
0.20
0.40
0.37
0.43
0.30
6
0.14
0.25
0.48
0.55
0.05
7
0.06
0.02
0.29
0.35
0.02
8
0.17
0.32
0.53
0.60
0.06
9
0.28
1.01
0.51
0.60
0.45
10
0
0
1.62
1.20
8.32 × 10 −6
11
0
0
2.53
1.33
6.61 × 10 −8
12
0
0
3.54
1.39
6.11 × 10 −10 (Table 4 continues on next page)
8
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom TABLE 4: Image Quality Evaluation Score on Visibility of Renal Parenchyma (continued) Scores for AIDR-Reconstructed Images Group, Image No.
Mean
SD
Scores for FBP-Reconstructed Images Mean
SD
p for ANOVA
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group D 1
0.42
0.54
0.38
0.64
0.88
2
0.49
0.52
0.23
0.45
0.18
3
0.63
0.72
0.38
0.73
0.38
4
0.51
0.65
0.12
0.23
0.04
5
0.62
0.76
0.33
0.69
0.29
6
0.44
0.60
0.40
0.62
0.88
7
0.80
0.72
0.22
0.59
0.03
8
1.00
0.87
0.09
0.34
8.94 × 10 −4
9
1.06
0.86
0.02
0.09
4.78 × 10 −5
10
0.92
0.92
0.35
0.80
11
2.32
1.29
0
0
1.55 × 10 −7
0.10
12
2.72
1.24
0
0
7.60 × 10 –9
13
3.91
1.16
0
0
4.89 × 10 −12
0.53
0.56
0.15
0.33
Group E 1
0.04
2
0.58
0.74
0.16
0.30
0.05
3
0.61
0.68
0.13
0.32
0.02
4
0.68
0.76
0.16
0.39
5
1.00
0.69
0
0
3.41 × 10 −6
0.03
6
0.84
0.69
0.06
0.15
2.76 × 10 −4
7
1.11
0.81
0.10
0.32
2.27 × 10 −4
8
1.75
0.97
0.02
0.07
2.90 × 10 −7
9
1.49
0.87
0
0
2.93 × 10 −7
10
2.77
1.32
0
0
1.41 × 10 −8
11
3.46
1.21
0
0
8.87 × 10 −11
12
4.21
0.85
0
0
4.92 × 10 −15
Note—AIDR = adaptive iterative dose reduction, ANOVA = analysis of variance, FBP = filtered back projection.
Acknowledgments We thank Hideji Otani, Ayumi Nitta-Seko, Yoko Murakami, Shobu Watanabe, and Yuki Tomozawa of our department for participating in the image-reading sessions. We also thank Noritoshi Ushio for technical assistance. References 1. Spielmann AL. Liver imaging with MDCT and high concentration contrast media. Eur J Radiol 2003; 45(suppl 1):S50–S52 2. Mortele KJ, McTavish J, Ros PR. Current techniques of computed tomography: helical CT, multidetector CT, and 3D reconstruction. Clin Liver Dis 2002; 6:29–52 3. Kopp AF, Heuschmid M, Claussen CD. Multidetector helical CT of the liver for tumor detection and characterization. Eur Radiol 2002; 12:745–752
4. Miéville FA, Gudinchet F, Rizzo E, et al. Paediatric cardiac CT examinations: impact of the iterative reconstruction method ASIR on image quality—preliminary findings. Pediatr Radiol 2011; 41:1154–1164 5. Ren Q, Dewan SK, Li M, et al. Comparison of adaptive statistical iterative and filtered back projection reconstruction techniques in brain CT. Eur J Radiol 2012; 81:2597–2601 6. Prakash P, Kalra MK, Digumarthy SR, et al. Radiation dose reduction with chest computed tomography using adaptive statistical iterative reconstruction technique: initial experience. J Comput Assist Tomogr 2010; 34:40–45 7. Vorona GA, Zuccoli G, Sutcavage T, Clayton BL, Ceschin RC, Panigrahy A. The use of adaptive statistical iterative reconstruction in pediatric head CT: a feasibility study. AJNR 2013; 34:205–211 8. Desai GS, Uppot RN, Yu EW, Kambadakone AR,
Sahani DV. Impact of iterative reconstruction on image quality and radiation dose in multidetector CT of large body size adults. Eur Radiol 2012; 22:1631–1640 9. Rapalino O, Kamalian S, Payabvash S, et al. Cranial CT with adaptive statistical iterative reconstruction: improved image quality with concomitant radiation dose reduction. AJNR 2012; 33:609–615 10. Ohno Y, Takenaka D, Kanda T, et al. Adaptive iterative dose reduction using 3D processing for reduced- and low-dose pulmonary CT: comparison with standard-dose CT for image noise reduction and radiological findings. AJR 2012; 199:[web] W477–W485. 11. Tatsugami F, Matsuki M, Nakai G, et al. The effect of adaptive iterative dose reduction on image quality in 320-detector row CT coronary angiography. Br J Radiol 2012; 85:e378–e382
AJR:202, January 2014 9
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Nitta et al. 12. Yamada Y, Jinzaki M, Hosokawa T, et al. Dose reduction in chest CT: comparison of the adaptive iterative dose reduction 3D, adaptive iterative dose reduction, and filtered back projection reconstruction techniques. Eur J Radiol 2012; 81:4185–4195 13. International Commission on Radiological Protection. Managing patient dose in computed tomography: ICRP publication 87. Ann ICRP 2000; 30 14. Tabachnick BG, Fidell LS. Computer-assisted research design and analysis. Needham Heights, MA: Allyn & Bacon, 2001:243–319 15. Singh S, Kalra MK, Gilman MD, et al. Adaptive statistical iterative reconstruction technique for radiation dose reduction in chest CT: a pilot study. Radiology 2011; 259:565–573 16. Rampado O, Bossi L, Garabello D, Davini O, Ropolo R. Characterization of a computed tomography iterative reconstruction algorithm by image quality evaluations with an anthropomorphic phantom. Eur J Radiol 2012; 81:3172–3177 17. Sagara Y, Hara AK, Pavlicek W, et al. Comparison of low-dose CT with adaptive statistical iterative reconstruction and routine-dose CT with filtered back
projection in 53 patients. AJR 2010; 195:713–719 18. Leipsic J, Labounty TM, Heilbron B, et al. Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography. AJR 2010; 195:649–654 19. Vorona GA, Ceschin RC, Clayton BL, Sutcavage T, Tadros SS, Panigrahy A. Reducing abdominal CT radiation dose with the adaptive statistical iterative reconstruction technique in children: a feasibility study. Pediatr Radiol 2011; 41:1174–1182 20. Flicek KT, Hara AK, Silva AC, Wu Q, Peter MB, Johnson CD. Reducing the radiation dose for CT colonography using adaptive statistical iterative reconstruction: a pilot study. AJR 2010; 195:126–131 21. Sato J, Akahane M, Inano S, et al. Effect of radiation dose and adaptive statistical iterative reconstruction on image quality of pulmonary computed tomography. Jpn J Radiol 2012; 30:146–153 22. Lee SH, Kim MJ, Yoon CS, Lee MJ. Radiation dose reduction with the adaptive statistical iterative reconstruction (ASIR) technique for chest CT in children: an intra-individual comparison. Eur J Radiol 2012; 81:e938–e943
23. Singh S, Kalra MK, Shenoy-Bhangle AS, et al. Radiation dose reduction with hybrid iterative reconstruction for pediatric CT. Radiology 2012; 263:537–546 24. Kim K, Kim YH, Kim SY, et al. Low-dose abdominal CT for evaluating suspected appendicitis. N Engl J Med 2012; 366:1596–1605 25. Priola AM, Priola SM, Giaj-Levra M, et al. Clinical implications and added costs of incidental findings in an early detection study of lung cancer by using low-dose spiral computed tomography. Clin Lung Cancer 2013; 14:139–148 26. Brady Z, Ramanauskas F, Cain TM, Johnston PN. Assessment of paediatric CT dose indicators for the purpose of optimisation. Br J Radiol 2012; 85:1488–1498 27. Namimoto T, Oda S, Utsunomiya D, et al. Improvement of image quality at low-radiation dose and lowcontrast material dose abdominal CT in patients with cirrhosis: intraindividual comparison of low tube voltage with iterative reconstruction algorithm and standard tube voltage. J Comput Assist Tomogr 2012; 36:495–501
Fig. 1—Upper abdominal phantom used in this study. A and B, Shown are photograph (A) and CT image (B) of phantom acquired at CT dose index of 13.1 mGy and reconstructed by filtered backprojection. It is possible to differentiate splenic vein from intrahepatic blood vessels and pancreatic parenchyma and to recognize surrounding area of inferior vena cava and renal parenchyma.
(Figures continue on next page)
10
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom Group B
Group A
Group C
3.0
2
3
3
0.8
4 5
5
0.7
6
6
0.6
8
0.5
10
0.4
12
2.5
7 9 11 13 14
0.3
4.0
1
2
2 3
4
2.0
4
7 8 9 10
1.5
11 12 13
1.0
5
3.0
14
6 7
2.5
8 9 10
2.0
11 12
1.5
13 14
1.0
0.2
0.5
0.5
0.1 0.0
1
3.5
Mean Dose (mGy)
1
0.9
Mean Dose (mGy)
Mean Dose (mGy)
AIDR
0.0
FBP
AIDR
0.0
FBP
A
C
4.5
1 2
3.5
4
Mean Dose (mGy)
7 8 9 10
2.0
11 12
1.5
3 4 5 6
6
2.5
2
3.5
5
3.0
1
4.0
3
13 14
3.0
8 9
2.0
11
0.5
0.5 0.0
D
10 12 13
1.0
FBP
7
2.5
1.5 1.0
AIDR
FBP
Group E
4.0
0.0
AIDR
B
Group D
Mean Dose (mGy)
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
1.0
14
AIDR
FBP
E
Fig. 2—Mean scores for images of phantom reconstructed with adaptive iterative dose reduction (AIDR) and filtered backprojection (FBP). A–E, Graphs show image quality scores assigned to each of 64 image pairs for visibility of renal parenchyma. Different lines and symbols denote image pairs. In all but highest-dose pair, in 13 group A pairs, visibility of renal parenchyma was scored as better on AIDR 3D images than on FBP images. This difference was statistically significant in two low-dose pairs. In contrast, in five group B and all group C pairs, visibility of renal parenchyma was judged to be better on FBP images than on AIDR 3D images. This difference was statistically significant in two and three low-dose pairs of groups B and C, respectively. For all image pairs of groups D and E, visibility of renal parenchyma was scored as better on AIDR 3D images than on FBP images. Difference was statistically significant in four low-dose pairs of group D and one standard-dose and five low-dose pairs of group E. Similar results were obtained for other evaluation criteria. Paired Student t test was used in this evaluation; level of significance was 0.000111607 (0.05/448).
AJR:202, January 2014 11
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Nitta et al.
A
B
C
D
Fig. 3—CT images of phantom. A–D, Images were reconstructed by adaptive iterative dose reduction (AIDR) 3D (A and C) and filtered backprojection (FBP) (B and D) at CT dose indexes of 2.5 mGy (A and B) and 4.5 mGy (C and D). Note lower level of noise in AIDR 3D images than in FBP images.
12
AJR:202, January 2014
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Special Articles Original Research
Norihisa Nitta1 Mitsuru Ikeda2 Akinaga Sonoda1 Yukihiro Nagatani1 Shinichi Ohta1 Masashi Takahashi1 Kiyoshi Murata1 Nitta N, Ikeda M, Sonoda A, et al.
Images Acquired Using 320-MDCT With Adaptive Iterative Dose Reduction With Wide-Volume Acquisition: Visual Evaluation of Image Quality by 10 Radiologists Using an Abdominal Phantom OBJECTIVE. The purpose of this study is to assess visual evaluations of CT images and to determine by how much radiation exposure dose could be reduced without compromising the image quality. MATERIALS AND METHODS. An abdominal CT phantom was scanned at 14 different tube currents. Raw data were reconstructed with adaptive iterative dose reduction (AIDR) 3D and filtered backprojection (FBP). We divided 64 different image pairs into five groups. Group A consisted of 14 image pairs acquired with AIDR 3D and FBP, groups B and D consisted of 13 pairs with a one-level exposure dose decrease in AIDR 3D and FBP, respectively, and groups C and E consisted of 12 pairs with a two-level exposure dose decrease in AIDR 3D and FBP, respectively. Ten radiologists participated in the reading session. Statistical analyses were calculated with analysis of variance and the paired Student t test. RESULTS. Analysis of variance of six criteria revealed that the results were better in groups A, D, and E when AIDR 3D was applied. Better results were obtained with FBP in groups B and C. When we subjected evaluations of the renal parenchyma to the Student t test, we found that the assigned scores were better with AIDR 3D in groups A, D, and E and better with FBP in groups B and C. Similar results were obtained for the other evaluation criteria. CONCLUSION. Visual subjective evaluation showed that images of acceptable quality could be obtained at dose reductions of approximately 10% in the high-dose range and about 20% in the moderate-dose range.
A
Keywords: 320-MDCT, abdominal phantom, filtered backprojection, image quality, iterative dose reduction DOI:10.2214/AJR.12.10364 Received December 1, 2012; accepted after revision April 3, 2013. 1 Department of Radiology, Shiga University of Medical Science, Tsukinowa-cho Seta, Otsu Shiga 520-2192, Japan. Address correspondence to N. Nitta. 2 Division of Radiological Technology, School of Health Sciences Radiological Technology, Nagoya University, Nagoya Aichi, Japan.
AJR 2014; 202:2–12 0361–803X/14/2021–2 © American Roentgen Ray Society
2
lthough MDCT studies yield clinically useful images, the radiation exposure dose tends to increase with diagnostic accuracy [1–4]. Because radiation exposure at CT may increase the risk for cancer in populations of a specific age and sex [1], it must be reduced without affecting image quality. To this end, manufacturers of CT scanners have developed iterative reconstruction techniques [5, 6], such as AIDR (adaptive iterative dose reduction; Toshiba Medical Systems), ASiR (adaptive statistical iterative reconstruction; GE Healthcare), iDose4 (Philips Healthcare), and SAFIRE (sinogram-affirmed iterative reconstruction; Siemens Healthcare). Although exploratory studies have shown that iterative reconstruction techniques facilitate a 20–30% dose reduction without compromising image quality [7–9], detailed data are lacking. Therefore, we used paired upper abdominal phantom images to compare the quality of iterative reconstruction CT images obtained at various radiation exposure doses
and of CT images acquired with filtered backprojection (FBP) at the same, higher, and lower doses. Materials and Methods Abdominal Phantom We used a commercially available upper abdominal CT phantom (PH-5, Kyoto Kagaku) consisting of bone and soft tissue (Fig. 1). The bone tissue is composed of epoxy resin, calcium carbonate (for the part corresponding to the medullary bone), and barium sulfate (for cortical bone). The soft tissue is made of hydroxyapatite and polyurethane resin whose mixture rate is adjusted to yield the same CT number (in Hounsfield units) as the corresponding abdominal organ. The longest and shortest diameter of the phantom are 26 and 18 cm, respectively, and the abdominal girth is 79 cm. The CT numbers of images of this phantom are 400 HU for vertebrae and ribs; 70 HU for the liver parenchyma; 50 HU for the spleen; 40 HU for portal and hepatic veins, the aorta, inferior vena cava, and other blood vessels; 30 HU for the pancreas and kidneys; 20 HU for the gallbladder; and −7 HU for soft tissue around organs.
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
CT Image Acquisition Using an abdominal CT phantom and a 320MDCT scanner (Aquillon ONE, Toshiba Medical Systems), we acquired images reconstructed with AIDR 3D (a modified AIDR technique developed by Toshiba Medical Systems) and with conventional FBP [10]. AIDR is a technology to selectively extract and remove only noise other than signals on the 3D voxels of reconstructed images [11]. AIDR 3D efficiently removes noise and artifacts from raw data using a statistical noise and scanner model. It extracts only noise and repeatedly removes noise at image reconstruction with an anatomic model [12]. We applied 14 levels of tube current, ranging from 20 to 550 mA. Other scanning parameters were tube voltage of 120 kV, collimation of 0.5 mm, scan time of 0.5 seconds, reconstruction kernel of FC04, and slice thickness of 5 mm. Automatic exposure control was not used in this study. In one rotation, we acquired CT images of the abdominal phantom (width, 16 cm) from the liver beneath the diaphragm to the lower pole of the kidney. We scanned each slice once at each scanning condition to obtain image pairs; one image was reconstructed with AIDR 3D and the other with the conventional FBP method. The reconstructed images were transferred to an imageviewing workstation (Ziostation, Ziosoft). The radiation absorption dose (exposure dose) was evaluated by CT dose index displayed on the console of the CT system. At our institute, the CT dose index is set at 20 mGy for routine abdominal CT examinations [13]. For convenience, we considered CT dose index at, above, and below approximately 20 mGy as moderate-, high-, and lowexposure radiation dose ranges, respectively.
Noise Evaluation To measure the image noise, we placed a circular 4-cm 2 region of interest (ROI) on five sites on each of 14 phantom CT images of the liver. ROIs were placed so that they were evaluated at exactly the same site in all 14 CT series by using a copyand-paste feature. These were acquired at varying tube currents, and the SD of the CT number in each ROI was calculated. We refer to this value as the noise evaluation.
Image Evaluation We prepared 64 image pairs; one CT image in each pair was reconstructed with AIDR 3D, and the corresponding image was reconstructed with FBP. Then we divided the image pairs into five groups as shown in Table 1. Group A contained 14 basic image pairs that were acquired at the same radiation exposure dose, ranging from 1.0 to 29.2 mGy. In groups B (13 pairs) and C (12 pairs), the
radiation exposure for the AIDR 3D–reconstructed image was one and two levels lower, respectively, than for the FBP image. In groups D (13 pairs) and E (12 pairs), the radiation exposure for the AIDR 3D–reconstructed image was one and two levels higher, respectively, than for the FBP image. The image pairs were displayed in landscape orientation on two high-resolution 24-inch monitors (MultiSync LCD2490WUXi2, NEC) placed side-by-side. The display conditions on the monitors were identical. The 64 image pairs were interpreted by 10 board-certified radiologists with 9–18 years of experience; they were blinded to the reconstruction methods and acquisition parameters. All readers understood the purpose of the study. The readers independently performed direct sideby-side comparisons of each CT image pair. Differences with respect to our seven evaluation criteria (visibility of intrahepatic vessels, discrimination between pancreas and splenic vein, visibility of the pancreatic margin, visibility of the margin between the inferior vena cava and aorta, strength of streak artifact from the vertebra, visibility of the renal parenchyma, and overall judgment) were recorded in pencil on a continuous 20-cm-long line corresponding to a Likert scale ranging from −5 to 5 (left and right end of the line, respectively), where −5 to −3 means that the left image is better, −3 to −1 means that the left image is slightly better, −1 to 1 means that the images are almost equivalent, 1 to 3 means that the right image is slightly better, and 3 to 5 means that the right image is better. Because the same image pair was evaluated twice with the display side changed, each observer read a total of 128 image pairs. These were displayed in random order; no two image pairs were displayed successively in the reading order. At each side-by-side comparison, the absolute value of the rating scale was assigned to the better image with a value of 0 assigned to the image on the other side. For image interpretation, the window and level were fixed at 350 and 40 HU, respectively. Viewing time was unlimited.
Statistical Analysis Excluding overall judgment, there were six evaluation criteria. These were defined as criteria A. To compare the evaluation scores of image pairs obtained with the two reconstruction methods, we performed analysis of variance (ANOVA) for a two repeated-measures design, using the obtained scores as the dependent variable, the two reconstruction methods as the within-subjects factor, and criteria A as the between-subjects factor [14]. If the null hypothesis in the Mauchly test of sphericity was rejected, we adjusted the df of F test statistics using the Greenhouse-Geisser test, the Huynh-Feldt test, and the lower-bound epsilon. For the statistical tests,
we adopted a significance level of 0.05. For this series of ANOVA tests, the adopted level of significance was 0.00078125 (0.05/64) to compensate for the multiple comparisons problem by the Bonferroni correction. For the interaction effect test, we did not use the Bonferroni correction; rather, we adopted a level of significance of 0.05 to decrease type 2 errors and to increase the statistical power. We also performed paired Student t tests to compare the evaluation scores for each of the six evaluation criteria plus overall judgment (criteria B) that were assigned to the image pairs reconstructed with the two different methods. To compensate for the multiple comparisons problem by the Bonferroni correction, we adopted a level of significance of 0.000111607 (0.05/448) for this series of paired Student t tests. To evaluate differences in the evaluation scores of images displayed on the right and left sides, we subjected all images, each image pair, and each criterion of criteria B to the paired Student t test. For this series of paired Student t tests, we did not use the Bonferroni correction and adopted a level of significance of 0.05 to decrease type 2 errors and to increase the statistical power. SSPS software (version 17, SPSS) was used for all statistical analyses.
Results On all 14 image pairs obtained at a dose ranging from 1.0 to 29.2 mGy, the noise evaluation of the AIDR 3D–reconstructed image was lower than that of the corresponding FBPreconstructed image. The maximum noise reduction was approximately 65% (Table 2). Table 3 shows the mean quality scores assigned for criteria A to each of the image pairs in the five groups. In all group A pairs, the quality of the AIDR 3D image was rated as better than that of the corresponding FBP image. The difference was statistically significant in four low-dose and two moderatedose pairs. Because, with the exception of one image pair, ANOVA revealed no statistically significant interaction between the reconstruction method and the evaluation criterion, this statistical difference was almost independent of the evaluation criterion except for the one image pair. On the other hand, in nine group B and 11 group C pairs, the quality of the AIDR 3D image was lower than that of the FBP image; this difference was statistically significant in three low-dose pairs of group B and all pairs of group C (Table 4). There was no statistically significant interaction between the reconstruction method and the evaluation criterion. In 11 group D and all 12 group E pairs, the quality of the AIDR 3D image was higher than
AJR:202, January 2014 3
Nitta et al. TABLE 1: CT Dose Index (CTDI), Tube Current, and Radiation Dose Reduction Rate of 64 Image Pairs AIDR-Reconstructed CT Image Group, Image Pair
CTDI (mGy)
Tube Current (mA)
FBP-Reconstructed CT Image CTDI (mGy)
Tube Current (mA)
Dose Reduction Rate (%)
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
29.2
550
29.2
550
0.0
2
26.6
500
26.6
500
0.0
3
24.4
460
24.4
460
0.0
4
22.3
420
22.3
420
0.0
5
20.2
380
20.2
380
0.0
6
18.1
340
18.1
340
0.0
7
15.9
300
15.9
300
0.0
8
13.1
260
13.1
260
0.0 0.0
9
11.1
220
11.1
220
10
9.1
180
9.1
180
0.0
11
7.0
140
7.0
140
0.0
12
4.5
90
4.5
90
0.0
13
2.5
50
2.5
50
0.0
14
1.0
20
1.0
20
0.0
Group B 1
26.6
500
29.2
550
8.9
2
24.4
460
26.6
500
8.3 8.6
3
22.3
420
24.4
460
4
20.2
380
22.3
420
9.4
5
18.1
340
20.2
380
10.4
6
15.9
300
18.1
340
12.2
7
13.1
260
15.9
300
17.6
8
11.1
220
13.1
260
15.3
9
9.1
180
11.1
220
18.0
10
7.0
140
9.1
180
19.1
11
4.5
90
7.0
140
35.7
12
2.5
50
4.5
90
44.4
13
1.0
20
2.5
50
60.0
1
24.4
460
29.2
550
16.4
2
22.3
420
26.6
500
16.2
3
20.2
380
24.4
460
17.2
4
18.1
340
22.3
420
18.9
Group C
5
15.9
300
20.2
380
21.3
6
13.1
260
18.1
340
27.6
7
11.1
220
15.9
300
30.2
8
9.1
180
13.1
260
30.5
9
7.0
140
11.1
220
36.9
10
4.5
90
9.1
180
50.1
11
2.5
50
7.0
140
64.3
12
1.0
20
4.5
90
77.8
Note—The reconstruction methods used in groups B and C were reversed in groups D and E. AIDR = adaptive iterative dose reduction, FBP = filtered backprojection.
4
AJR:202, January 2014
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Quality of 320-MDCT Images of Abdominal Phantom that of the FBP image. This difference was statistically significant in the eight group D and 11 group E image pairs. There was no statistically significant interaction between the reconstruction method and the evaluation criterion. For all evaluation criteria except overall judgment (criteria B), the assigned scores were similar to the mean scores. Figure 2 is an illustrative example. It shows the score assigned for visibility of the renal parenchyma to each of the 64 image pairs. Table 4 shows the mean quality scores assigned for the renal parenchyma to each of the image pairs in the five groups. With the exception of the highest-dose image pair, in 13 group A pairs, visibility of the renal parenchyma was scored as better on AIDR 3D images than on FBP images. This difference was statistically significant in the two low-dose pairs. On the other hand, in five group B and all group C pairs, visibility of the renal parenchyma was judged to be better on FBP images than on AIDR 3D images. This difference was statistically significant in two and three low-dose pairs of groups B and C, respectively. For all image pairs of groups D and E, visibility of the renal parenchyma was scored to be better on AIDR 3D images than on FBP images. The difference was statistically significant in four low-dose pairs of group D and one moderate-dose and five low-dose pairs of group E. Similar results were obtained for the other evaluation criteria. In fact, the quality of the image displayed on the left side was judged to be significantly better than that on the right side (p < 0.000001) (Fig. 3). There was a statistically significant difference in the quality score assigned to images displayed on the left and right sides in 55 of the total 896 evaluations (64 images × 2 sides × 7 criteria = 896 evaluations), although we detected no particular trend to explain this observation. In some cases, the image quality scores assigned to each side were identical. Each image pair was evaluated twice with the display side reversed to compensate for bias. Because our search for potential factors contributing to this bias detected none, we think that the bias is incidental. Discussion Our study shows that, by the subjective judgment of 10 readers, the quality of CT images reconstructed with AIDR 3D was better than that of FBP-reconstructed images obtained at the same radiation exposure dose. This difference in quality was more evident at lower radiation doses. On the other hand,
TABLE 2: Noise Evaluation of 14 Phantom CT Images Image No.
Radiation Exposure Dose (mGy)
1 2
Image Noise (Mean ± SD), Reconstruction Method
Tube Current (mA)
AIDR
FBP
29.2
550
5.80 ± 0.48
8.70 ± 0.58
26.6
500
6.22 ± 0.40
8.62 ± 1.12
3
24.4
460
6.56 ± 0.53
9.62 ± 0.66
4
22.3
420
6.68 ± 0.55
9.78 ± 0.50 10.36 ± 0.62
5
20.2
380
7.00 ± 0.52
6
18.1
340
7.52 ± 0.89
11.08 ± 1.01
7
15.9
300
8.10 ± 0.64
12.04 ± 0.79
8
13.1
260
8.44 ± 0.85
12.82 ± 1.17
9
11.1
220
8.70 ± 0.52
13.58 ± 0.82
10
9.1
180
9.68 ± 0.74
15.42 ± 1.23
11
7.0
140
10.52 ± 0.74
17.38 ± 1.74
12
4.5
90
12.48 ± 0.70
22.36 ± 1.74
13
2.5
50
17.22 ± 1.80
33.64 ± 3.64
14
1.0
20
25.56 ± 1.68
63.78 ± 6.87
Note—Noise evaluation was calculated from the SD of the CT numbers in each 4-cm2 circular region of interest placed at five sites on the liver in each phantom CT image. AIDR = adaptive iterative dose reduction, FBP = filtered backprojection.
there were no instances in which the subjectively judged quality of AIDR 3D images was significantly better than that of FBP images obtained at the higher dose, and there were instances in which the quality of AIDR 3D images was judged to be statistically worse than that of FBP images obtained at the higher radiation dose. Consequently, we cannot adduce evidence for the contention that the AIDR 3D method allows a reduction in the radiation dose without a detrimental effect on the image quality. If the probability of the null hypothesis for a statistical test is posited to be above 0.1, then the quality of image pairs is almost equivalent for the statistical test. Under this hypothesis, our ANOVA results for criteria A suggest that the subjectively judged quality of members of image pairs 1, 2, 3, 5, and 7 of group B and of the members of pair 1 in group C is almost equivalent (Table 3). In other words, with AIDR 3D, a dose reduction of 9–17% can be obtained without rendering the subjective quality of images worse than that of images acquired at high and standard radiation doses. The results of ANOVA for criteria A also indicate that the subjective image quality can be considered to be almost equivalent in image pairs 1, 2, and 5 of group D in Table 3. That is, the quality of AIDR 3D CT images acquired at the high and moderate radiation
dose will also be almost equivalent to that of FBP images obtained with a dose reduction of 9–10%. On the basis of these considerations, we suggest that the subjective image quality evaluation of CT images obtained at the high dose level is robust. According to the noise evaluation, the noise level of FBP-reconstructed CT images scanned at 29.2, 22.3, 13.1, and 7.0 mGy was almost equivalent to that of AIDR 3D–reconstructed images scanned at 11.1, 9.1, 4.5, and 2.5 mGy, respectively. This finding suggests that, compared with FBP, AIDR 3D reduces the radiation exposure dose by 60–65% at image acquisition based solely on the noise level. However, contrary to our expectations, we found that, in almost all instances, the quality of AIDR 3D images obtained with a dose reduction above 25% was subjectively judged to be worse than that of FBP images. This difference was statistically significant in all cases where the dose reduction exceeded 60%. Thus, a dose reduction of 9–17% may be feasible at upper abdominal CT performed at the standard- and high-dose level when the images are reconstructed with AIDR 3D. In studies on tolerable dose reduction rates for iterative reconstruction techniques, the reduction rate was higher [15, 16]. Although two or three readers participated in most other studies on the subjective quality of CT images reconstructed with iterative reconstruction techniques [15,
AJR:202, January 2014 5
Nitta et al. TABLE 3: Image Quality Evaluation Scores on Six Evaluation Criteria Excluding Overall Judgment (Criteria A) Scores for AIDR-Reconstructed Images Group, Image
Mean
Scores for FBP-Reconstructed Images
SD
Mean
p
SD
ANOVA
Interaction
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
0.48
0.73
0.38
0.75
0.40
0.91
2
0.56
0.68
0.27
0.50
0.0018
0.33
3
0.49
0.73
0.26
0.53
0.02
0.94
4
0.48
0.73
0.31
0.58
0.10
0.80
5
0.50
0.65
0.28
0.65
0.03
0.71
6
0.57
0.77
0.21
0.47
0.0002
0.98
7
0.47
0.57
0.16
0.37
0.00003
0.96
8
0.41
0.58
0.26
0.47
0.06
0.78 0.92
9
0.65
0.83
0.25
0.80
0.001
10
0.61
0.84
0.26
0.58
0.002
0.78
11
0.61
0.59
0.11
0.29
1.50 × 10 −10
0.89
12
0.73
0.92
0.22
0.48
1.18 × 10 −5
0.86
13
0.88
0.73
0.02
0.13
0.00
0.03
14
1.20
1.06
0.004
0.04
0.00
0.23
Group B 1
0.35
0.55
0.30
0.48
0.51
0.56
2
0.33
0.59
0.28
0.40
0.49
0.14
3
0.31
0.47
0.37
0.67
0.50
0.95
4
0.15
0.27
0.38
0.62
0.001
0.55
5
0.32
0.52
0.32
0.51
0.94
0.91
6
0.23
0.46
0.32
0.48
0.19
0.32
7
0.38
0.57
0.35
0.59
0.74
0.76
8
0.24
0.43
0.45
0.58
0.009
0.09
9
0.26
0.45
0.46
0.54
0.02
0.96
10
0.35
0.50
0.29
0.45
0.42
0.53
11
0.11
0.33
1.00
0.89
6.6 × 10 −16
0.17
12
0.02
0.11
1.27
1.14
0.00
0.61
13
0.01
0.06
2.52
1.59
0.00
0.61
1
0.30
0.58
0.22
0.38
0.31
0.67
2
0.09
0.20
0.43
0.54
4.27 × 10 −8
0.15 0.19
Group C
3
0.08
0.33
0.79
0.90
5.06 × 10 −12
4
0.17
0.38
0.42
0.52
4.00 × 10 −4
0.45
5
0.12
0.31
0.45
0.49
2.70 × 10 −7
0.07
6
0.12
0.24
0.45
0.57
8.69 × 10 −7
0.19 0.14
7
0.07
0.19
0.33
0.43
7.55 × 10 −8
8
0.08
0.24
0.70
0.76
1.08 × 10 −12
0.08
9
0.22
0.72
0.68
0.80
5.87 × 10 −5
0.65
10
0.004
0.05
1.56
1.18
0.00
0.75
11
0.02
0.13
2.45
1.36
0.00
0.86
12
0.00
0.00
3.34
1.41
0.00
0.94 (Table 3 continues on next page)
6
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom TABLE 3: Image Quality Evaluation Scores on Six Evaluation Criteria Excluding Overall Judgment (Criteria A) (continued) Scores for AIDR-Reconstructed Images Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group, Image
Scores for FBP-Reconstructed Images
p
Mean
SD
Mean
SD
ANOVA
Interaction
0.29
0.55
0.39
0.65
0.26
0.66
Group D 1 2
0.42
0.59
0.32
0.59
0.26
0.26
3
0.56
0.74
0.33
0.56
0.03
0.99
4
0.43
0.63
0.13
0.26
4.53 × 10 −5
0.93
5
0.44
0.65
0.38
0.69
0.58
0.73
6
0.36
0.59
0.46
0.71
0.35
0.86
7
0.60
0.71
0.26
0.57
7.40 × 10 −4
0.82 0.79 0.24
8
0.89
0.81
0.12
0.38
8.82 × 10 −13
9
0.94
0.87
0.06
0.20
0.00
10
0.84
0.99
0.25
0.59
5.35 × 10 −6
0.98
11
2.07
1.09
0.002
0.02
0.00
0.90
12
2.52
1.29
0.04
0.45
0.00
0.26
13
3.70
1.23
0.00
0.00
0.00
0.95
1
0.43
0.61
0.21
0.42
8.00 × 10 −3
0.78
2
0.45
0.62
0.17
0.30
3.00 × 10 −4
0.66
3
0.51
0.67
0.21
0.40
4.00 × 10 −4
0.22
4
0.56
0.72
0.19
0.45
6.41 × 10 −5
0.99
Group E
5
0.83
0.65
0.03
0.15
0.00
0.57
6
0.88
0.86
0.08
0.22
3.11 × 10 −15
0.25
7
0.88
0.87
0.09
0.25
5.45 × 10 −14
0.73
8
1.67
1.11
0.03
0.10
0.00
0.93
9
1.39
0.90
0.02
0.14
0.00
0.49
10
2.55
1.25
0.003
0.04
0.00
0.67
11
3.30
1.24
0.00
0.00
0.00
0.82
12
4.08
0.94
0.02
0.18
0.00
0.84
Note—AIDR = adaptive iterative dose reduction, ANOVA = analysis of variance, FBP = filtered backprojection.
17–21], 10 experienced radiologists evaluated our phantom CT images independently. Consequently, the power of our statistical tests is higher than that of the other studies. In addition, we performed direct comparisons by displaying image pairs side by side. We think that the dose reduction rate that is possible with iterative reconstruction techniques without affecting the image quality is lower than has been reported in the above-cited studies. Because our readers performed subjective evaluations of the quality of images displayed side by side, we cannot rule out the possibility of display-side bias. In fact, images displayed on the left side received significantly better evaluation scores. To compensate for this bias, each reader evaluated each
image pair twice with the display side of the images reversed. Our study has some limitations. First, training data were not provided to the radiologists at the image reading sessions and only the image evaluation method was explained. Second, only unenhanced phantom CT images were evaluated, and clinical and contrast-enhanced CT images must be included in future evaluation studies. Third, the range of the radiation doses we were able to use was limited and was determined by our CT system. Fourth, as in other studies on the quality of CT images reconstructed with iterative reconstruction techniques [8, 22, 23], we did not investigate the image quality necessary for a clinical diagnosis. We ex-
pect that the radiation dose reduction without a significant effect on the diagnosis will be greater than the dose reduction that does not affect the subjective evaluation of image quality [24–27]. Conclusion At the same radiation exposure dose, the subjectively judged quality of phantom upper abdominal CT images reconstructed with AIDR 3D was higher than that of images reconstructed with FBP. By use of AIDR 3D reconstruction, the dose for upper abdominal CT scans can be reduced by 9–17% of the standard- and high-dose level without degradation of the subjective image quality.
AJR:202, January 2014 7
Nitta et al. TABLE 4: Image Quality Evaluation Score on Visibility of Renal Parenchyma Scores for AIDR-Reconstructed Images Group, Image No.
Mean
SD
Scores for FBP-Reconstructed Images Mean
SD
p for ANOVA
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group A 1
0.42
0.60
0.50
0.99
0.79
2
0.69
0.82
0.21
0.45
0.06
3
0.58
0.72
0.25
0.51
0.18
4
0.55
0.90
0.31
0.50
0.38
5
0.54
0.66
0.11
0.21
0.02
6
0.66
0.80
0.14
0.38
0.03
7
0.52
0.63
0.19
0.42
0.11
8
0.53
0.69
0.15
0.30
0.06
9
0.70
0.83
0.23
0.40
0.07
10
0.85
0.88
0.21
0.51
0.03
11
0.70
0.56
0.09
0.27
1.00 × 10 −3
12
1.02
1.13
0.22
0.46
13
1.22
0.89
0
0
7.49 × 10 −6
0.02
14
1.39
1.26
0
0
9.25 × 10 −5
Group B 1
0.44
0.57
0.19
0.30
0.15
2
0.37
0.53
0.30
0.45
0.73
3
0.28
0.47
0.31
0.64
0.88
4
0.16
0.26
0.30
0.44
0.33
5
0.37
0.55
0.37
0.52
1.00
6
0.32
0.50
0.17
0.37
0.36
7
0.44
0.54
0.21
0.42
0.22
8
0.29
0.40
0.22
0.34
0.65
9
0.32
0.51
0.45
0.54
0.55
10
0.40
0.58
0.36
0.48
0.87
11
0.22
0.46
0.86
0.70
0.01
12
0
0
1.14
0.97
4.70 × 10 −5
13
0
0
2.72
1.52
1.67 × 10 −7
Group C 1
0.24
0.48
0.28
0.41
0.81
2
0.06
0.17
0.56
0.59
3.00 × 10 −3
3
0.11
0.42
0.65
0.81
0.02
4
0.22
0.43
0.39
0.41
0.31
5
0.20
0.40
0.37
0.43
0.30
6
0.14
0.25
0.48
0.55
0.05
7
0.06
0.02
0.29
0.35
0.02
8
0.17
0.32
0.53
0.60
0.06
9
0.28
1.01
0.51
0.60
0.45
10
0
0
1.62
1.20
8.32 × 10 −6
11
0
0
2.53
1.33
6.61 × 10 −8
12
0
0
3.54
1.39
6.11 × 10 −10 (Table 4 continues on next page)
8
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom TABLE 4: Image Quality Evaluation Score on Visibility of Renal Parenchyma (continued) Scores for AIDR-Reconstructed Images Group, Image No.
Mean
SD
Scores for FBP-Reconstructed Images Mean
SD
p for ANOVA
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Group D 1
0.42
0.54
0.38
0.64
0.88
2
0.49
0.52
0.23
0.45
0.18
3
0.63
0.72
0.38
0.73
0.38
4
0.51
0.65
0.12
0.23
0.04
5
0.62
0.76
0.33
0.69
0.29
6
0.44
0.60
0.40
0.62
0.88
7
0.80
0.72
0.22
0.59
0.03
8
1.00
0.87
0.09
0.34
8.94 × 10 −4
9
1.06
0.86
0.02
0.09
4.78 × 10 −5
10
0.92
0.92
0.35
0.80
11
2.32
1.29
0
0
1.55 × 10 −7
0.10
12
2.72
1.24
0
0
7.60 × 10 –9
13
3.91
1.16
0
0
4.89 × 10 −12
0.53
0.56
0.15
0.33
Group E 1
0.04
2
0.58
0.74
0.16
0.30
0.05
3
0.61
0.68
0.13
0.32
0.02
4
0.68
0.76
0.16
0.39
5
1.00
0.69
0
0
3.41 × 10 −6
0.03
6
0.84
0.69
0.06
0.15
2.76 × 10 −4
7
1.11
0.81
0.10
0.32
2.27 × 10 −4
8
1.75
0.97
0.02
0.07
2.90 × 10 −7
9
1.49
0.87
0
0
2.93 × 10 −7
10
2.77
1.32
0
0
1.41 × 10 −8
11
3.46
1.21
0
0
8.87 × 10 −11
12
4.21
0.85
0
0
4.92 × 10 −15
Note—AIDR = adaptive iterative dose reduction, ANOVA = analysis of variance, FBP = filtered back projection.
Acknowledgments We thank Hideji Otani, Ayumi Nitta-Seko, Yoko Murakami, Shobu Watanabe, and Yuki Tomozawa of our department for participating in the image-reading sessions. We also thank Noritoshi Ushio for technical assistance. References 1. Spielmann AL. Liver imaging with MDCT and high concentration contrast media. Eur J Radiol 2003; 45(suppl 1):S50–S52 2. Mortele KJ, McTavish J, Ros PR. Current techniques of computed tomography: helical CT, multidetector CT, and 3D reconstruction. Clin Liver Dis 2002; 6:29–52 3. Kopp AF, Heuschmid M, Claussen CD. Multidetector helical CT of the liver for tumor detection and characterization. Eur Radiol 2002; 12:745–752
4. Miéville FA, Gudinchet F, Rizzo E, et al. Paediatric cardiac CT examinations: impact of the iterative reconstruction method ASIR on image quality—preliminary findings. Pediatr Radiol 2011; 41:1154–1164 5. Ren Q, Dewan SK, Li M, et al. Comparison of adaptive statistical iterative and filtered back projection reconstruction techniques in brain CT. Eur J Radiol 2012; 81:2597–2601 6. Prakash P, Kalra MK, Digumarthy SR, et al. Radiation dose reduction with chest computed tomography using adaptive statistical iterative reconstruction technique: initial experience. J Comput Assist Tomogr 2010; 34:40–45 7. Vorona GA, Zuccoli G, Sutcavage T, Clayton BL, Ceschin RC, Panigrahy A. The use of adaptive statistical iterative reconstruction in pediatric head CT: a feasibility study. AJNR 2013; 34:205–211 8. Desai GS, Uppot RN, Yu EW, Kambadakone AR,
Sahani DV. Impact of iterative reconstruction on image quality and radiation dose in multidetector CT of large body size adults. Eur Radiol 2012; 22:1631–1640 9. Rapalino O, Kamalian S, Payabvash S, et al. Cranial CT with adaptive statistical iterative reconstruction: improved image quality with concomitant radiation dose reduction. AJNR 2012; 33:609–615 10. Ohno Y, Takenaka D, Kanda T, et al. Adaptive iterative dose reduction using 3D processing for reduced- and low-dose pulmonary CT: comparison with standard-dose CT for image noise reduction and radiological findings. AJR 2012; 199:[web] W477–W485. 11. Tatsugami F, Matsuki M, Nakai G, et al. The effect of adaptive iterative dose reduction on image quality in 320-detector row CT coronary angiography. Br J Radiol 2012; 85:e378–e382
AJR:202, January 2014 9
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Nitta et al. 12. Yamada Y, Jinzaki M, Hosokawa T, et al. Dose reduction in chest CT: comparison of the adaptive iterative dose reduction 3D, adaptive iterative dose reduction, and filtered back projection reconstruction techniques. Eur J Radiol 2012; 81:4185–4195 13. International Commission on Radiological Protection. Managing patient dose in computed tomography: ICRP publication 87. Ann ICRP 2000; 30 14. Tabachnick BG, Fidell LS. Computer-assisted research design and analysis. Needham Heights, MA: Allyn & Bacon, 2001:243–319 15. Singh S, Kalra MK, Gilman MD, et al. Adaptive statistical iterative reconstruction technique for radiation dose reduction in chest CT: a pilot study. Radiology 2011; 259:565–573 16. Rampado O, Bossi L, Garabello D, Davini O, Ropolo R. Characterization of a computed tomography iterative reconstruction algorithm by image quality evaluations with an anthropomorphic phantom. Eur J Radiol 2012; 81:3172–3177 17. Sagara Y, Hara AK, Pavlicek W, et al. Comparison of low-dose CT with adaptive statistical iterative reconstruction and routine-dose CT with filtered back
projection in 53 patients. AJR 2010; 195:713–719 18. Leipsic J, Labounty TM, Heilbron B, et al. Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography. AJR 2010; 195:649–654 19. Vorona GA, Ceschin RC, Clayton BL, Sutcavage T, Tadros SS, Panigrahy A. Reducing abdominal CT radiation dose with the adaptive statistical iterative reconstruction technique in children: a feasibility study. Pediatr Radiol 2011; 41:1174–1182 20. Flicek KT, Hara AK, Silva AC, Wu Q, Peter MB, Johnson CD. Reducing the radiation dose for CT colonography using adaptive statistical iterative reconstruction: a pilot study. AJR 2010; 195:126–131 21. Sato J, Akahane M, Inano S, et al. Effect of radiation dose and adaptive statistical iterative reconstruction on image quality of pulmonary computed tomography. Jpn J Radiol 2012; 30:146–153 22. Lee SH, Kim MJ, Yoon CS, Lee MJ. Radiation dose reduction with the adaptive statistical iterative reconstruction (ASIR) technique for chest CT in children: an intra-individual comparison. Eur J Radiol 2012; 81:e938–e943
23. Singh S, Kalra MK, Shenoy-Bhangle AS, et al. Radiation dose reduction with hybrid iterative reconstruction for pediatric CT. Radiology 2012; 263:537–546 24. Kim K, Kim YH, Kim SY, et al. Low-dose abdominal CT for evaluating suspected appendicitis. N Engl J Med 2012; 366:1596–1605 25. Priola AM, Priola SM, Giaj-Levra M, et al. Clinical implications and added costs of incidental findings in an early detection study of lung cancer by using low-dose spiral computed tomography. Clin Lung Cancer 2013; 14:139–148 26. Brady Z, Ramanauskas F, Cain TM, Johnston PN. Assessment of paediatric CT dose indicators for the purpose of optimisation. Br J Radiol 2012; 85:1488–1498 27. Namimoto T, Oda S, Utsunomiya D, et al. Improvement of image quality at low-radiation dose and lowcontrast material dose abdominal CT in patients with cirrhosis: intraindividual comparison of low tube voltage with iterative reconstruction algorithm and standard tube voltage. J Comput Assist Tomogr 2012; 36:495–501
Fig. 1—Upper abdominal phantom used in this study. A and B, Shown are photograph (A) and CT image (B) of phantom acquired at CT dose index of 13.1 mGy and reconstructed by filtered backprojection. It is possible to differentiate splenic vein from intrahepatic blood vessels and pancreatic parenchyma and to recognize surrounding area of inferior vena cava and renal parenchyma.
(Figures continue on next page)
10
AJR:202, January 2014
Quality of 320-MDCT Images of Abdominal Phantom Group B
Group A
Group C
3.0
2
3
3
0.8
4 5
5
0.7
6
6
0.6
8
0.5
10
0.4
12
2.5
7 9 11 13 14
0.3
4.0
1
2
2 3
4
2.0
4
7 8 9 10
1.5
11 12 13
1.0
5
3.0
14
6 7
2.5
8 9 10
2.0
11 12
1.5
13 14
1.0
0.2
0.5
0.5
0.1 0.0
1
3.5
Mean Dose (mGy)
1
0.9
Mean Dose (mGy)
Mean Dose (mGy)
AIDR
0.0
FBP
AIDR
0.0
FBP
A
C
4.5
1 2
3.5
4
Mean Dose (mGy)
7 8 9 10
2.0
11 12
1.5
3 4 5 6
6
2.5
2
3.5
5
3.0
1
4.0
3
13 14
3.0
8 9
2.0
11
0.5
0.5 0.0
D
10 12 13
1.0
FBP
7
2.5
1.5 1.0
AIDR
FBP
Group E
4.0
0.0
AIDR
B
Group D
Mean Dose (mGy)
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
1.0
14
AIDR
FBP
E
Fig. 2—Mean scores for images of phantom reconstructed with adaptive iterative dose reduction (AIDR) and filtered backprojection (FBP). A–E, Graphs show image quality scores assigned to each of 64 image pairs for visibility of renal parenchyma. Different lines and symbols denote image pairs. In all but highest-dose pair, in 13 group A pairs, visibility of renal parenchyma was scored as better on AIDR 3D images than on FBP images. This difference was statistically significant in two low-dose pairs. In contrast, in five group B and all group C pairs, visibility of renal parenchyma was judged to be better on FBP images than on AIDR 3D images. This difference was statistically significant in two and three low-dose pairs of groups B and C, respectively. For all image pairs of groups D and E, visibility of renal parenchyma was scored as better on AIDR 3D images than on FBP images. Difference was statistically significant in four low-dose pairs of group D and one standard-dose and five low-dose pairs of group E. Similar results were obtained for other evaluation criteria. Paired Student t test was used in this evaluation; level of significance was 0.000111607 (0.05/448).
AJR:202, January 2014 11
Downloaded from www.ajronline.org by Univeristy of Brighton on 07/04/14 from IP address 194.81.203.94. Copyright ARRS. For personal use only; all rights reserved
Nitta et al.
A
B
C
D
Fig. 3—CT images of phantom. A–D, Images were reconstructed by adaptive iterative dose reduction (AIDR) 3D (A and C) and filtered backprojection (FBP) (B and D) at CT dose indexes of 2.5 mGy (A and B) and 4.5 mGy (C and D). Note lower level of noise in AIDR 3D images than in FBP images.
12
AJR:202, January 2014
