- Email: [email protected]
Noise-based tube current reduction method with iterative reconstruction for reduction of radiation exposure in coronary CT angiography
European Journal of Radiology 82 (2013) 349–355
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Noise-based tube current reduction method with iterative reconstruction for reduction of radiation exposure in coronary CT angiography Junlin Shen a,1 , Xiangying Du a,1 , Daode Guo a,1 , Lizhen Cao a,1 , Yan Gao a,1 , Mei Bai b,2 , Pengyu Li a,1 , Jiabin Liu a,1 , Kuncheng Li a,∗ a b
Department of Radiology, Xuanwu Hospital of Capital Medical University, No. 45, Chang-Chun St., Xicheng District, Beijing 100053, China Department of Biomedical Engineering, Xuanwu Hospital of Capital Medical University, No. 45, Chang-Chun St., Xicheng District, Beijing 100053, China
a r t i c l e
i n f o
Article history: Received 6 August 2012 Received in revised form 6 October 2012 Accepted 8 October 2012 Keywords: Iterative reconstruction Coronary CT angiography Radiation dose
a b s t r a c t Purpose: To investigate the potential of noise-based tube current reduction method with iterative reconstruction to reduce radiation exposure while achieving consistent image quality in coronary CT angiography (CCTA). Materials and methods: 294 patients underwent CCTA on a 64-detector row CT equipped with iterative reconstruction. 102 patients with fixed tube current were assigned to Group 1, which was used to establish noise-based tube current modulation formulas, where tube current was modulated by the noise of test bolus image. 192 patients with noise-based tube current were randomly assigned to Group 2 and Group 3. Filtered back projection was applied for Group 2 and iterative reconstruction for Group 3. Qualitative image quality was assessed with a 5 point score. Image noise, signal intensity, volume CT dose index, and dose-length product were measured. Results: The noise-based tube current modulation formulas were established through regression analysis using image noise measurements in Group 1. Image noise was precisely maintained at the target value of 35.00 HU with small interquartile ranges for Group 2 (34.17–35.08 HU) and Group 3 (34.34–35.03 HU), while it was from 28.41 to 36.49 HU for Group 1. All images in the three groups were acceptable for diagnosis. A relative 14% and 41% reduction in effective dose for Group 2 and Group 3 were observed compared with Group 1. Conclusion: Adequate image quality could be maintained at a desired and consistent noise level with overall 14% dose reduction using noise-based tube current reduction method. The use of iterative reconstruction further achieved approximately 40% reduction in effective dose. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Coronary CT angiography (CCTA) has emerged as a useful diagnostic tool for the noninvasive assessment of coronary artery disease with excellent sensitivity and good specificity. However, radiation dose for CCTA draw much attention because of the usage of sub-millimeter slice thickness and small helical pitches associated with retrospective ECG gating. Currently, various
∗ Corresponding author. Tel.: +86 10 83198376; fax: +86 10 83198376. E-mail addresses: [email protected] (J. Shen), [email protected] (X. Du), [email protected] (D. Guo), [email protected] (L. Cao), [email protected] (Y. Gao), [email protected] (M. Bai), [email protected] (P. Li), [email protected] (J. Liu), [email protected] (K. Li). 1 Tel: +86 10 83198376; fax: +86 10 83198376. 2 Tel: +86 10 83198353; fax: +86 10 83198353. 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.10.008
technological innovations, such as tube voltage reduction [1], ECG gated tube current modulation [2], prospective ECG gating [3], high-pitch helical scanning [4], spiral dynamic or axial adaptive z-collimation, and wide-volume acquisition [5] are available to lower radiation exposure during CCTA. Even so, dose reduction remains limited by the use of filter back projection (FBP) reconstruction, which produces a significant increase of image noise in the case of excessive reduction of radiation exposure [6]. Alternative technique of image reconstruction with iterative reconstruction (IR) has become available for clinical use in recent years. By iterative comparison of each synthesized forward projection to the actual measurements and modeling system statistics, IR has the potential to selectively reduce image noise, which may permit preserved image quality with reduced tube current [7]. However, dose-saving technologies above including IR cannot compensate for the chest attenuation variations among different patients. Therefore, modulation of tube current individually
350
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
is essential for radiation dose optimization. With noise-based tube current modulation, the noise measured on test bolus image directly reflects the chest attenuation in each patient and can be used to optimize radiation exposure while achieving consistent image quality for CCTA [8]. IR algorithm and noise-based tube current reduction method are different approaches, but they can be complementary and may have an additive effect on dose reduction. However, to our knowledge, no studies have formally integrated the IR algorithm to the noisebased tube current reduction method, and no studies have focused on the dose reduction performance of IR compared with FBP at the same image noise level individually in a large patient cohort. Accordingly, the purpose of this prospective study was to investigate the potential of noise-based tube current reduction method with IR algorithm to reduce radiation exposure while achieving consistent image quality over a wide range of patient population. 2. Materials and methods This study consists of two parts: the establishment of noisebased tube current modulation formulas with FBP reconstruction and IR, respectively; the investigation of noise-based tube current reduction method with IR to reduce radiation exposure while achieving consistent image quality. The study protocol had been approved by the local ethics committee. Written informed consent was obtained from each patient. Prospective ECG gating is possible on our scanner that offers great possibilities for dose reduction. However, retrospectively gated CCTA was specifically request by the ordering physician for assessment of left ventricular function. 2.1. Establishment of noise-based tube current reduction method 2.1.1. Study population From October 2011 to November 2011, a total of 117 consecutive patients scheduled for CCTA were prospectively recruited. Exclusion criteria were pregnant and lactating women, known contradiction to iodinated contrast agent, inability to sustain a breath hold in the allotted time, acute coronary syndrome, heart failure, heart rate higher than 70 bpm, and irregular cardiac rhythm. Thus, 11 patients were excluded and 106 patients were eligible for CCTA. Of the 106 patients eligible for CCTA, 102 patients were successfully scanned with fixed tube current (Group 1: 44 male, 58 female, age from 34 to 81 with mean of 57.5 years, BMI from 18.9 to 34.6 with mean of 25.5 kg/m2 ). 2.1.2. Scan protocol and image reconstruction All examinations were performed on a 64-detector row CT equipped with IR technique (Discovery CT750 HD, GE Healthcare, Milwaukee, Wisconsin, USA). The sequences consisted of scout scan for positioning of the heart, calcium scoring scan for minimum scan length, test bolus scan for delay time and CCTA scan for coronary angiography and ventricular function. 20 ml iodinated contrast (Ultravist 370; Schering, Berlin, Germany) was injected intravenously followed by 20 ml saline at a rate of 5 ml/s for test bolus, while 60 ml iodinated contrast was injected followed by 40 ml saline at the rate of 5 ml/s for CCTA. The test bolus scan was obtained using 1 × 5 mm collimation, axial scan mode, 100 kV, 60 mA, 1 s rotation speed, FOV of 25 mm × 25 mm, 512 × 512 matrix size and slice thickness of 5 mm. The retrospectively gated CCTA parameters were as following: 64 × 0.625 mm collimation, cardiac helical scan mode, 0.35 s rotation speed, FOV of 25 mm × 25 mm, 512 × 512 matrix size, 0.625 mm slice thickness and 0.18-0.24 helical pitch. Tube voltage was fixed at 120 kVp and 100 kVp was not used. With the goal of minimizing radiation exposure, ECG gated tube current modulation was used where the peak tube current was applied for the CCTA phases and 20% of the peak value was applied
for the rest of the cardiac phases. When heart rate was ≤65 bpm, the CCTA phases were 70–80%. When heart rate was at 65–70 bpm, the CCTA phases were 40–80%. Peak tube current was fixed at 600 mA for all patients in Group 1, and was patient-dependent based on the noise of test bolus image in Group 2 and Group 3 Images were reconstructed from CCTA raw data with both FBP and IR and two sets of images were produced. IR applied in this study was ASIR (Adaptive Statistical Iterative Reconstruction; GE Healthcare, Milwaukee, Wisconsin, USA), which was performed from both the projection data and image data. Only a limited number of iterations are required to complete an entire analysis for ASIR, so the average reconstruction time was approximately 40–60% longer with the ASIR than with the standard FBP. In our study, IR represents a composite of 40% IR and 60% FBP [9]. 2.1.3. Noise-based tube current reduction method Image noise is defined as the standard deviation (SD) of CT attenuation values over a relatively uniform area. SD is a first approximation used for estimation of noise, while in reality this function is more complex. The image noise measurement was performed by placing a circular region of interest (ROI) of about 100 mm2 in the center of aortic root on CCTA image (SDCCTA ) and test bolus image (SDTB ), respectively. Simple linear regression analysis was conducted to assess the relationship of SDCCTA as a function of SDTB , and an equation was to be obtained: SDCCTA = b × SDTB + a, where SDCCTA is the CCTA image noise with a fixed tube current. The simple linear regression model above shows that the expectation of the dependent variable SDCCTA is linear with the independent variable SDTB , with an intercept a and a slop b. Since image noise is known to be inversely proportional to the square root of the radiation exposure (or tube current) [10], we obtained the following equation:
SD
CCTA
SD0
2
=
mACCTA 600
where 600 is the value of the fixed tube current (mA) used in our CCTA scanning; SD0 is the desired CCTA image noise level; mACCTA is the required noise-based tube current to get SD0 . Combining the two equations above, we obtained the relationship between mACCTA and SDTB (noise-based tube current modulation formulas): mACCTA = 600 ×
(a × SD + b) 2 TB SD0
Finally, noise-based tube current modulation formulas with and without IR were to be established from two sets of images, respectively. These two formulas were used to predict the required tube current to obtain the desired CCTA image noise based on the test bolus noise measurement. We chose 35 HU as the desired noise level (SD0 ) for CCTA from the long term experience of our institute, since it was supposed enough for quantification of stenosis. 2.2. Clinical investigation of noise-based tube current reduction method with iterative reconstruction 2.2.1. Study population From December 2011 to February 2012, a total of 219 consecutive patients scheduled for CCTA were prospectively recruited. Exclusion criteria in this part were the same as the previous part and 20 patients were excluded. 199 patients eligible for CCTA were assigned into noise-based tube current modulation protocol with FBP or IR by random number table. 99 patient with FBP (Group 2:
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Fig. 1. Flow chart of patient inclusion.
351
CCTA: coronary CT angiography; FBP: filter back projection; IR: iterative reconstruction.
48 male, 51 female, age from 37 to 82 with mean of 59.0 years, BMI from 17.9 to 33.3 with mean of 25.5 kg/m2 ) and 93 patent with IR (Group 3: 53 male, 41 female, age from 32 to 83 with mean of 59.6 years, BMI from 16.4 to 35.8 with mean of 25.2 kg/m2 ) were successfully scanned.
2.2.2. Scan protocol and image reconstruction Patients in Group 2 and Group 3 following the same scan parameters as Group 1, except the full tube current being individually calculated with noise-based tube current modulation formulas based on the noise measured from the test bolus image. A flow chart (Fig. 1) illustrates the 3 different acquisition protocols used in our study. Since not all tube current settings were available, we choose the tube current closest to the suggested value.
2.2.3. Qualitative image quality CCTA image data were transferred to an offline workstation (AW 4.5 Advantage, GE Healthcare, Milwaukee, Wisconsin, USA) and independently assessed by two cardiovascular radiologists (Pengyu Li and Jiabin Liu, each with 6 years of experience) who were blinded to technical and clinical information. Curved multiplanar reformations were assessed for each of the four main arteries (left main, left anterior descending, left circumflex and right coronary artery) of each patient. Signal intensity at the aortic root was chosen as
the new window center whereas the window width was arbitrarily defined between +1000 HU and +2000 HU. According to the diagnostic acceptability and the Likert score [9], a 5-point score (Fig. 2) assigned on the basis of the worst scored artery was used: 1 = impaired image quality limited by excessive lumen noise and poor vessel wall definition or poor lumen attenuation, unacceptable for diagnosis; 2 = reduced image quality with massive lumen noise and poor vessel wall definition or low lumen attenuation, diagnosis acceptable only under limited conditions; 3 = adequate image quality with moderate lumen noise and minimal limitation of vessel wall definition, acceptable for diagnosis; 4 = good image quality with minimal lumen noise and well maintained vessel wall definition, fully acceptable for diagnosis; and 5 = excellent image quality with clear vessel wall definition and limited perceptible lumen noise, fully acceptable for diagnosis. Any artery was considered acceptable for diagnosis if image quality was adequate for quantification of luminal stenosis with a minimum diameter of 1.5 mm. Segments with stent graft and extensive calcifications were excluded from analysis. The readers were instructed to ignore motion artifact and stair-step artifact. The disagreement between the two readers was settled by a consensus reading. 2.2.4. Quantitative image quality The measurement of signal intensity (MeanCCTA ) and image noise (SDCCTA ) were performed by placing a circular ROI of about
352
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Table 1 Noise-based tube current modulation formulas regarding SDTB and mACCTA. Group
SDTB (HU)
mACCTA (mA)
Fixed tube current with FBP (Group 1) Noise-based tube current with FBP (Group 2) Noise-based tube current with IR (Group 3)
19.48 (17.07–22.72)
600
P value
19.00 (16.88–21.71)
600 × [(9.550 + 1.177SDTB )/35]2
<0.0001
1.006–1.347
0.808
19.07 (16.40–22.21)
600 × [(8.365 + 0.856SDTB )/35]2
<0.0001
0.722–0.989
0.786
–
95% CI –
r value –
Note: SDTB are median with interquartile range (IQR). SDTB : timing bolus noise; mACCTA : tube current of CCTA; 95%CI: 95% confidence interval for confidence.
100 mm2 in the center of aortic root at the level of the left main artery. MeanCCTA was derived from the CT attenuation values on axial source CCTA images. SDCCTA was defined as the standard deviations of CT attenuation values on axial source CCTA images. Signal-to-noise ratio was calculated as MeanCCTA divided by SDCCTA . All Quantitative image quality was measured by one experienced reader (Lizhen Cao, 9 years of experience). 2.2.5. Radiation exposure Radiation exposure for CCTA was determined by volume CT dose index (CTDIvol), dose-length product (DLP), and effective dose. The DLP was converted to mSv by multiplying it by a chest-specific conversion coefficient of 0.014 mSv/(mGy cm) [11]. 2.2.6. Statistical analysis Statistical analysis was performed using SPSS software (18.0, SPSS Inc., Chicago, III). A P value of less than 0.05 was considered statistically significant difference. Quantitative variables were expressed as mean ± SD and median with interquartile range (IQR) as appropriate; categorical variables were expressed as counts (or proportions in percent). Ordinal variables (i.e., image quality) were also given in means ± SD. The differences between the three groups were assessed with one-way ANOVA for normally distributed variables, and Kruskal–Wallis test for non-normally distributed quantitative and
ordinal variables. For categorical variables, we use chi-square test. For subgroup analysis, a Bonferroni adjustment was made for multiple comparisons. Interobserver agreement for qualitative image quality was estimate using kappa statistic.
3. Results 3.1. Noise-based tube current reduction method Simple linear regression analysis using image noise measurements in Group 1 resulted in equation relating SDCCTA to SDTB . The inverse square relationship between image noise and radiation dose resulted in equation relating SDCCTA to mACCTA . Then, combining the two equations above, noise-based tube current modulation formulas for Group2 and Group3 were established (Table 1). With the noise-based tube current reduction method, the required mACCTA to obtain desired image noise level for CCTA image can be obtained from SDTB on test bolus image. We chose 35 HU as the desired noise level (SD0 ) for CCTA. 3.2. Patient and scan characteristics A total of 294 patients were successfully scanned without adverse events. Comparisons of patient and scan characteristics among three groups are listed in Table 2. No significant differences were observed in terms of patient and scan characteristics, except for tube current, which was lowest for Group 3 (median, 299 mA; IQR, 244–364 mA) and highest for Group 1 (600 mA).
3.3. Qualitative and quantitative image quality
Fig. 2. Quantitative image quality measurement. Region of interest (circle) in the center of aortic root at the level of the left main coronary artery illustrating measurements of signal intensity (434 HU) and image noise (35 HU), as perfromed in every patient.
Interobserver agreement for image quality between the 2 readers was good ( = 0.76). The mean image quality score was 3.38 ± 0.73 for Group 1, which was significantly higher than Group 2 (3.11 ± 0.35, p < 0.05) and Group 3 (3.10 ± 0.33, p < 0.05) (Table 3). However, both Group 2 and Group 3 were scored adequate image quality and similar image quality score was found between the two groups (Fig. 3). Notably, the standard deviation of the image quality score in Group 2 and Group 3 was smaller than that of the Group 1, reflecting more consistent image quality obtained in Group 2 and Group 3. Quantitative analysis demonstrated that no significant difference in signal intensity was observed among three groups, while image noise was significantly lower for Group 1 compared with Group 2 and Group 3. Median image noise was precisely maintained at 35.00 for both Group 2 and Group 3. The IQR of the image noise in Group 2 and Group 3 was much smaller than that of Group 1, reflecting more consistent image noise obtained in Group 2 and Group 3. Signal-to-noise ratio for Group 1 (13.1 ± 3.8) was higher compared with Group 2 (11.9 ± 2.2; p = 0.013) and Group 3 (12.21; p = 0.036). However, Group 3 and Group 2 were comparable regarding signalto-noise ratio (p = 1.000).
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
353
Table 2 Patient and scan characteristics. Characteristics
Fixed tube current with FBP (Group 1)
Noise-based tube current with FBP (Group 2)
Noise-based tube current with IR (Group 3)
n Age, years Male sex Heart ratea , beat/min Height, m Weight, kg Body mass index, kg/m2 Tube current, mA Scan length, mm Pitch Phase (70–80%)
102 57.7 ± 9.5 43.1 (44/102) 56.8 ± 5.6 1.65 ± 0.08 69.8 ± 12.1 25.5 ± 3.4 600 125.1 ± 13.2 0.22 ± 0.01 72.5 (74/102)
99 59.0 ± 9.0 48.5 (48/99) 56.3 ± 5.4 1.65 ± 0.08 69.2 ± 10.1 25.5 ± 2.9 484 (425–604) 127.2 ± 10.9 0.22 ± 0.02 73.7 (73/99)
93 59.6 ± 9.0 55.9 (52/93) 57.2 ± 6.2 1.66 ± 0.07 69.2 ± 10.3 25.2 ± 3.3 299 (244–364) 128.8 ± 12.1 0.22 ± 0.02 67.7 (63/93)
P value
0.248 0.203 0.562 0.624 0.902 0.704 <0.0001 0.113 0.486 0.625
Data are mean ± standard deviation (SD), median with interquartile range (IQR) and percentage with raw data in parentheses. a Heart rate is mean heart rate during scanning.
3.4. Radiation dose estimates CTDIvol was lowest for Group 3 (median, 30.0 mGy; IQR, 24.8–33.7 mGy), followed by Group 2 (median, 43.4 mGy; IQR, 37.3–53.3 mGy) and the by Group 1 (median, 51.0 mGy; IQR, 47.6–58.8 mGy; all p < 0.05). Similarly, the DLP was significant lower for Group 2 (median, 704 mGy cm; IQR, 627–828 mGy cm) compared to Group 1 (median, 822 mGy cm; IQR, 764–924 mGy cm), and further reduced by Group 3 (median, 487 mGy cm; IQR, 407–547 mGy cm; all p < 0.05) compared to Group 2. This corresponds to a relative 15% and 41% reduction in effective dose for Group 2 (median, 9.9 mSv; IQR, 8.8–11.6 mSv) and Group 3 (median, 6.8 mSv; IQR, 5.7–7.7 mSv) compared with Group 1 (median, 11.5 mSv; IQR, 10.7–12.9 mSv; both p < 0.05). That is, the effective dose can be further reduced by a relative 31% for Group 3 compared to Group 2 (p < 0.05).
4. Discussion In this prospective single center study, we established the noisebased tube current reduction methods with and without IR, and then compared protocols with and without noise-based tube current and that with and without IR concerning image quality and radiation exposure. We demonstrate that the use of the noisebased tube current modulation could reduce the effective dose by 14%, compared with fixed tube current protocol while maintaining clinically acceptable images with consistent image noise (35 HU) across the entire patient population, and the effective dose could be further reduced by 31% with IR algorithm. Radiation exposure is a major concern for CCTA especially in case of repeated examinations or in particular subgroups of patients, such as children and young women [1]. A worldwide radiation dose survey indicated that CCTA scan is associated with a median dose of 12 mSv [12]. Accordingly, the median radiation dose of 11.5 mSv
(IQR, 10.7–12.9 mSv) in the fixed tube current protocol of our study compares well with the previous study. Fixed tube current protocol, generally designed to obtain sufficient quality for all patients, leads to unnecessary radiation exposure in the majority of patients, in particular slim. Therefore, modulation of tube current individually is essential for radiation dose optimization. Tube current modulation according to body mass index or weight has been recognized as the standard method and yielded relative good result [13]. However, due to the difference in chest attenuation between men and women, and differences in fat distribution among individuals, the accuracy of this approach is limited [14,15]. In addition, some cardiac and pericardial diseases (e.g. hypertrophic cardiomyopathy, dilated cardiomyopathy and pericardial effusion) can also cause the inaccuracy of the approach above. Noise-based tube current reduction method used in our study modulates the tube current individually according to test bolus image. The noise measured on test bolus image directly reflects the chest attenuation in each patient. Therefore, more accurate tube current modulation and radiation dose optimization can be achieved for subsequent CCTA [8]. Our results show that tube current was significant lower for protocol with noise-based tube current than protocol with fixed tube current using the same FBP algorithm, leading to a significant reduction of effective dose by 14%. Thought radiation dose optimization with noise-based tube current modulation leads to significant dose reduction in our study, it does not necessary mean dose reduction for all patients; for some extreme large patients more radiation dose was required to produce the desired image quality. The relative reduction of effective dose in our study (14%) was lower than that of Qi et al. (28%) [8], which may be due to the statistical method used to establish the noise-based tube current formulas. Specifically, simple linear regression in this study was performed to assess the linear relationship of image noise of CCTA as a function of image noise of test bolus. This was different from the correlation analysis adopted by the study from Qi et al,
Table 3 Results on image quality and radiation exposure. Characteristics
Fixed tube current with FBP (Group 1)
Noise-based tube current with FBP (Group 2)
Noise-based tube current with IR (Group 3)
P value
Image quality score Signal intensity, HU Image noise, HU Signal-to-noise ratio CTDIvol, mGy DLP, mGy cm Effective dose, mSv
3.38 ± 0.73 416 ± 70 32.48 (28.41–36.49) 13.1 ± 3.8 51.0 (47.6–58.8) 822 (764–924) 11.5 (10.7–12.9)
3.11 ± 0.35 413 ± 70 35.00 (34.17–35.08) 11.9 ± 2.2 43.4 (37.3–53.3) 704 (627–828) 9.9 (8.8–11.6)
3.10 ± 0.33 417 ± 67 35.00 (34.34–35.03) 12.1 ± 2.1 30.0 (24.8–33.7) 487 (407–547) 6.8 (5.7–7.7)
<0.0001 0.884 0.001 0.007 <0.0001 <0.0001 <0.0001
Note: CTDIvol: volume CT dose index; DLP: dose-length product. Data are mean ± standard deviation (SD), median with interquartile range (IQR) in parentheses.
354
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Fig. 3. Representative CCTA demonstrating image quality with different scan protocols. 60 year old woman (body mass index, 25.68 kg/m2 ) (a–c) and 60 year old man (body mass index, 22.03 kg/m2 ) (d–f) both with chest pain. Curved multiplanar reformations of left descending arteries (a, d), left circumflex arteries (b, e), and right coronary arteries (c, f) of two patients demonstrating the same noise level (35 HU) and image quality (score 3) obtained by the noise-based tube current method with (374 mA, a–c) and without iterative reconstruction (189 mA, d–f).
where the purpose was to examine the strength and direction of the relationship between the two variables [16]. The noise-based tube current reduction method has been proposed to determine the required tube current for a desired image noise individually. However, the desired noise level has not been scientifically supported in CCTA and should be determined on the basis of the study purpose (e.g. ruling out significant stenosis, quantification of stenosis, or plaque characteristic analysis) [14]. In order to keep radiation dose as low as reasonably achievable
(ALARA), we chose 35 HU as the desired noise level for CCTA image from the long term experience of our institute, which was higher than previous studies of 27–28 HU [8,17], since it was supposed enough for quantification of stenosis. In consistent with the desired image noise level, the median image noise in this study was precisely maintained at 35 HU with a small amount of variability [(IQR, 34.17–35.08 HU) with FBP and (IQR, 34.34–35.03 HU) with IR], reflecting more consistent image noise achieved using noise-based tube current method. Though image quality score was
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
significantly lower with noise-based tube current method in this study, it was still adequate for diagnosis. More importantly, the consistent image noise achieved using noise-based tube current method leads to more uniform qualitative image quality than that with fixed tube current protocol. Despite the development of several technological innovations for CCTA within the past few years, dose reduction remains limited by the use of FBP algorithm, which produces a significant increase of image noise in the case of excessive reduction of radiation dose [6]. By iterative comparison of each synthesized forward projection to the actual measurements and modeling system statistics (photo statistics and electronic noise), IR has the potential to selectively reduce image noise, which may permit preserved image quality with reduced tube current, thereby permitting lower radiation dose [7]. In consistence with the previous study, IR was not associated with decrease in signal and SNR. In our study, noise-based tube current method with IR algorithm saved 41% effective dose compared with fixed tube current protocol using FBP algorithm. Though higher effective does reduction (44%) was observed between the two algorithms in the ERASIR study [18]. The difference in relative dose savings may be due to the other dose reduction techniques (i.e., sequential scan mode, 100 kV tube voltage) employed in the ERASIR study. Therefore, after adjustment for these scan settings, IR was only associated with a 27% reduction in effective dose in comparison to FBP. In addition, though no significant difference was observed in image noise in the ERASIR study, it was not intended to maintain at the same level for each patient and tube current was merely controlled based on experience of each study site. On contrast, our study made use of the noise-based method to modulate the tube current individually and maintain the median noise level precisely at 35 HU with small IQR, which effectively eliminating difference in chest attenuation and leading to the consistent image quality for each patient. Our study had several limitations. First, the radiation dose associated with the scout scan, calcium score scan, and test bolus scan was not assessed in this study. However, compared with CCTA, the radiation dose of these pre-control scans represents only a small fraction of the total dose. Second, the retrospectively gated CCTA was used in addition to assess left ventricular function. Compared to retrospectively gating technique, studies with prospective gating have reported a 69% effective dose reduction [19]. Noise-based tube current reduction method can also be applied for prospective gated CCTA. Future studies are warranted to investigate this method in prospective gated CCTA. Third, the IR algorithm used in our study represents a modified and limited form of IR. More complete IR can further improve image quality with substantial decreased image noise, allowing further reduction in radiation dose [20]. However, high computational cost and long reconstruction times remain a barrier for the complete IR in daily clinical application. Forth, radiation dose varied with the square of tube voltage in the setting of a constant tube current. Therefore, reducing tube voltage has a greater effect on the reduction of radiation exposure than reducing tube current. Radiation dose was relatively high in our study because the tube voltage was fixed at 120 kVp and only tube current was modulated. Future studies are warranted to integrate low tube voltage (100 kVp or 80 kVp) into the noise-based tube current modulation method. 5. Conclusions We firstly integrate IR algorithm into noise-based tube current method to optimize radiation dose in CCTA. Therefore, image noise could be precisely maintained at the desired level individually and consistent image quality could be obtained across the entire patient
355
population. At the same time, a 41% reduction in effective dose is achieved compared with a fixed tube current protocol. Conflict of interest No conflict of interest exits in the submission of this manuscript. Acknowledgment The authors would like to thank Dr Jianying Li, Huizhi Cao and Shichen Liu for technique suggestions and language revision. Our work was supported by the National Natural Science Foundation (30970821) and National Basic Research Program 973 (2010CB732600) of China. References [1] Hausleiter J, Martinoff S, Hadamitzky M, et al. Image quality and radiation exposure with a low tube voltage protocol for coronary CT angiography results of the PROTECTION II Trial. JACC: Cardiovascular Imaging 2010;3(11): 1113–23. [2] Jakobs TF, Becker CR, Ohnesorge B, et al. Multislice helical CT of the heart with retrospective ECG gating: reduction of radiation exposure by ECG-controlled tube current modulation. European Radiology 2002;12(5):1081–6. [3] Earls JP, Berman EL, Urban BA, et al. Prospectively gated transverse coronary CT angiography versus retrospectively gated helical technique: improved image quality and reduced radiation dose. Radiology 2008;246(3):742–53. [4] Alkadhi H, Stolzmann P, Desbiolles L, et al. Low-dose, 128-slice, dual-source CT coronary angiography: accuracy and radiation dose of the high-pitch and the step-and-shoot mode. Heart 2010;96(12):933–8. [5] Matthew JW, Mark EO, Milind YD, et al. New radiation dose saving technologies for 256-slice cardiac computed tomography angiography. International Journal of Cardiovascular Imaging 2009;25(Suppl. 2):189–99. [6] Gervaise A, Osemont B, Lecocq S, et al. CT image quality improvement using adaptive iterative dose reduction with wide-volume acquisition on 320detector CT. European Radiology 2012;22(2):295–301. [7] Hara AK, Paden RG, Silva AC, et al. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. American Journal of Roentgenology 2009;193(3):764–71. [8] Qi W, Li J, Du X. Method for automatic tube current selection for obtaining a consistent image quality and dose optimization in a cardiac multidetector CT. Korean Journal of Radiology 2009;10(6):568–74. [9] Leipsic J, Labounty TM, Heilbron B, et al. Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography. American Journal of Roentgenology 2010;195(3):649–54. [10] Boedeker KL, Mcnitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part II. Noise power spectra and signal to noise. Physics in Medicine and Biology 2007;52(14):4047–61. [11] American Association of Medical Physicists (AAPM) Task Group 23. CT dosimetry. The measurement, reporting and management of radiation dose in CT. Available at: http://www.aapm.org/pubs/reports (accessed 01.08.10). [12] Hausleiter J, Meyer T, Hermann F, et al. Estimated radiation dose associated with cardiac CT angiography. JAMA 2009;301(5):500–7. [13] Tatsugami F, Husmann L, Herzog BA, et al. Evaluation of a body mass index-adapted protocol for low-dose 64-MDCT coronary angiography with prospective ECG triggering. American Journal of Roentgenology 2009;192(3):635–8. [14] Hur G, Hong SW, Kim SY, et al. Uniform image quality achieved by tube current modulation using SD of attenuation in coronary CT angiography. American Journal of Roentgenology 2007;189(1):188–96. [15] Paul JF, Abada HT. Strategies for reduction of radiation dose in cardiac multislice CT. European Radiology 2007;17(8):2028–37. [16] Zou KH, Tuncali K, Silverman SG. Correlation and simple linear regression. Radiology 2003;227(3):617–28. [17] Gao J, Li J, Earls J, et al. Individualized tube current selection for 64-row cardiac CTA based on analysis of the scout view. European Journal of Radiology 2011;79(2):266–71. [18] Leipsic J, Labounty TM, Heilbron B, et al. Estimated radiation dose reduction using adaptive statistical iterative reconstruction in coronary CT angiography: the ERASIR study. American Journal of Roentgenology 2010;195(3): 655–60. [19] Hausleiter J, Meyer TS, Martuscelli E, et al. Image quality and radiation exposure with prospectively ECG-triggered axial scanning for coronary CT angiography: the multicenter, multivendor. randomized PROTECTION-III study. JACC: Cardiovascular Imaging 2012;5(5):484–93. [20] Scheffel H, Stolzmann P, Schlett CL, et al. Coronary artery plaques: cardiac CT with model-based and adaptive-statistical iterative reconstruction technique. European Journal of Radiology 2011;81(3):363–9.
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Noise-based tube current reduction method with iterative reconstruction for reduction of radiation exposure in coronary CT angiography Junlin Shen a,1 , Xiangying Du a,1 , Daode Guo a,1 , Lizhen Cao a,1 , Yan Gao a,1 , Mei Bai b,2 , Pengyu Li a,1 , Jiabin Liu a,1 , Kuncheng Li a,∗ a b
Department of Radiology, Xuanwu Hospital of Capital Medical University, No. 45, Chang-Chun St., Xicheng District, Beijing 100053, China Department of Biomedical Engineering, Xuanwu Hospital of Capital Medical University, No. 45, Chang-Chun St., Xicheng District, Beijing 100053, China
a r t i c l e
i n f o
Article history: Received 6 August 2012 Received in revised form 6 October 2012 Accepted 8 October 2012 Keywords: Iterative reconstruction Coronary CT angiography Radiation dose
a b s t r a c t Purpose: To investigate the potential of noise-based tube current reduction method with iterative reconstruction to reduce radiation exposure while achieving consistent image quality in coronary CT angiography (CCTA). Materials and methods: 294 patients underwent CCTA on a 64-detector row CT equipped with iterative reconstruction. 102 patients with fixed tube current were assigned to Group 1, which was used to establish noise-based tube current modulation formulas, where tube current was modulated by the noise of test bolus image. 192 patients with noise-based tube current were randomly assigned to Group 2 and Group 3. Filtered back projection was applied for Group 2 and iterative reconstruction for Group 3. Qualitative image quality was assessed with a 5 point score. Image noise, signal intensity, volume CT dose index, and dose-length product were measured. Results: The noise-based tube current modulation formulas were established through regression analysis using image noise measurements in Group 1. Image noise was precisely maintained at the target value of 35.00 HU with small interquartile ranges for Group 2 (34.17–35.08 HU) and Group 3 (34.34–35.03 HU), while it was from 28.41 to 36.49 HU for Group 1. All images in the three groups were acceptable for diagnosis. A relative 14% and 41% reduction in effective dose for Group 2 and Group 3 were observed compared with Group 1. Conclusion: Adequate image quality could be maintained at a desired and consistent noise level with overall 14% dose reduction using noise-based tube current reduction method. The use of iterative reconstruction further achieved approximately 40% reduction in effective dose. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Coronary CT angiography (CCTA) has emerged as a useful diagnostic tool for the noninvasive assessment of coronary artery disease with excellent sensitivity and good specificity. However, radiation dose for CCTA draw much attention because of the usage of sub-millimeter slice thickness and small helical pitches associated with retrospective ECG gating. Currently, various
∗ Corresponding author. Tel.: +86 10 83198376; fax: +86 10 83198376. E-mail addresses: [email protected] (J. Shen), [email protected] (X. Du), [email protected] (D. Guo), [email protected] (L. Cao), [email protected] (Y. Gao), [email protected] (M. Bai), [email protected] (P. Li), [email protected] (J. Liu), [email protected] (K. Li). 1 Tel: +86 10 83198376; fax: +86 10 83198376. 2 Tel: +86 10 83198353; fax: +86 10 83198353. 0720-048X/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2012.10.008
technological innovations, such as tube voltage reduction [1], ECG gated tube current modulation [2], prospective ECG gating [3], high-pitch helical scanning [4], spiral dynamic or axial adaptive z-collimation, and wide-volume acquisition [5] are available to lower radiation exposure during CCTA. Even so, dose reduction remains limited by the use of filter back projection (FBP) reconstruction, which produces a significant increase of image noise in the case of excessive reduction of radiation exposure [6]. Alternative technique of image reconstruction with iterative reconstruction (IR) has become available for clinical use in recent years. By iterative comparison of each synthesized forward projection to the actual measurements and modeling system statistics, IR has the potential to selectively reduce image noise, which may permit preserved image quality with reduced tube current [7]. However, dose-saving technologies above including IR cannot compensate for the chest attenuation variations among different patients. Therefore, modulation of tube current individually
350
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
is essential for radiation dose optimization. With noise-based tube current modulation, the noise measured on test bolus image directly reflects the chest attenuation in each patient and can be used to optimize radiation exposure while achieving consistent image quality for CCTA [8]. IR algorithm and noise-based tube current reduction method are different approaches, but they can be complementary and may have an additive effect on dose reduction. However, to our knowledge, no studies have formally integrated the IR algorithm to the noisebased tube current reduction method, and no studies have focused on the dose reduction performance of IR compared with FBP at the same image noise level individually in a large patient cohort. Accordingly, the purpose of this prospective study was to investigate the potential of noise-based tube current reduction method with IR algorithm to reduce radiation exposure while achieving consistent image quality over a wide range of patient population. 2. Materials and methods This study consists of two parts: the establishment of noisebased tube current modulation formulas with FBP reconstruction and IR, respectively; the investigation of noise-based tube current reduction method with IR to reduce radiation exposure while achieving consistent image quality. The study protocol had been approved by the local ethics committee. Written informed consent was obtained from each patient. Prospective ECG gating is possible on our scanner that offers great possibilities for dose reduction. However, retrospectively gated CCTA was specifically request by the ordering physician for assessment of left ventricular function. 2.1. Establishment of noise-based tube current reduction method 2.1.1. Study population From October 2011 to November 2011, a total of 117 consecutive patients scheduled for CCTA were prospectively recruited. Exclusion criteria were pregnant and lactating women, known contradiction to iodinated contrast agent, inability to sustain a breath hold in the allotted time, acute coronary syndrome, heart failure, heart rate higher than 70 bpm, and irregular cardiac rhythm. Thus, 11 patients were excluded and 106 patients were eligible for CCTA. Of the 106 patients eligible for CCTA, 102 patients were successfully scanned with fixed tube current (Group 1: 44 male, 58 female, age from 34 to 81 with mean of 57.5 years, BMI from 18.9 to 34.6 with mean of 25.5 kg/m2 ). 2.1.2. Scan protocol and image reconstruction All examinations were performed on a 64-detector row CT equipped with IR technique (Discovery CT750 HD, GE Healthcare, Milwaukee, Wisconsin, USA). The sequences consisted of scout scan for positioning of the heart, calcium scoring scan for minimum scan length, test bolus scan for delay time and CCTA scan for coronary angiography and ventricular function. 20 ml iodinated contrast (Ultravist 370; Schering, Berlin, Germany) was injected intravenously followed by 20 ml saline at a rate of 5 ml/s for test bolus, while 60 ml iodinated contrast was injected followed by 40 ml saline at the rate of 5 ml/s for CCTA. The test bolus scan was obtained using 1 × 5 mm collimation, axial scan mode, 100 kV, 60 mA, 1 s rotation speed, FOV of 25 mm × 25 mm, 512 × 512 matrix size and slice thickness of 5 mm. The retrospectively gated CCTA parameters were as following: 64 × 0.625 mm collimation, cardiac helical scan mode, 0.35 s rotation speed, FOV of 25 mm × 25 mm, 512 × 512 matrix size, 0.625 mm slice thickness and 0.18-0.24 helical pitch. Tube voltage was fixed at 120 kVp and 100 kVp was not used. With the goal of minimizing radiation exposure, ECG gated tube current modulation was used where the peak tube current was applied for the CCTA phases and 20% of the peak value was applied
for the rest of the cardiac phases. When heart rate was ≤65 bpm, the CCTA phases were 70–80%. When heart rate was at 65–70 bpm, the CCTA phases were 40–80%. Peak tube current was fixed at 600 mA for all patients in Group 1, and was patient-dependent based on the noise of test bolus image in Group 2 and Group 3 Images were reconstructed from CCTA raw data with both FBP and IR and two sets of images were produced. IR applied in this study was ASIR (Adaptive Statistical Iterative Reconstruction; GE Healthcare, Milwaukee, Wisconsin, USA), which was performed from both the projection data and image data. Only a limited number of iterations are required to complete an entire analysis for ASIR, so the average reconstruction time was approximately 40–60% longer with the ASIR than with the standard FBP. In our study, IR represents a composite of 40% IR and 60% FBP [9]. 2.1.3. Noise-based tube current reduction method Image noise is defined as the standard deviation (SD) of CT attenuation values over a relatively uniform area. SD is a first approximation used for estimation of noise, while in reality this function is more complex. The image noise measurement was performed by placing a circular region of interest (ROI) of about 100 mm2 in the center of aortic root on CCTA image (SDCCTA ) and test bolus image (SDTB ), respectively. Simple linear regression analysis was conducted to assess the relationship of SDCCTA as a function of SDTB , and an equation was to be obtained: SDCCTA = b × SDTB + a, where SDCCTA is the CCTA image noise with a fixed tube current. The simple linear regression model above shows that the expectation of the dependent variable SDCCTA is linear with the independent variable SDTB , with an intercept a and a slop b. Since image noise is known to be inversely proportional to the square root of the radiation exposure (or tube current) [10], we obtained the following equation:
SD
CCTA
SD0
2
=
mACCTA 600
where 600 is the value of the fixed tube current (mA) used in our CCTA scanning; SD0 is the desired CCTA image noise level; mACCTA is the required noise-based tube current to get SD0 . Combining the two equations above, we obtained the relationship between mACCTA and SDTB (noise-based tube current modulation formulas): mACCTA = 600 ×
(a × SD + b) 2 TB SD0
Finally, noise-based tube current modulation formulas with and without IR were to be established from two sets of images, respectively. These two formulas were used to predict the required tube current to obtain the desired CCTA image noise based on the test bolus noise measurement. We chose 35 HU as the desired noise level (SD0 ) for CCTA from the long term experience of our institute, since it was supposed enough for quantification of stenosis. 2.2. Clinical investigation of noise-based tube current reduction method with iterative reconstruction 2.2.1. Study population From December 2011 to February 2012, a total of 219 consecutive patients scheduled for CCTA were prospectively recruited. Exclusion criteria in this part were the same as the previous part and 20 patients were excluded. 199 patients eligible for CCTA were assigned into noise-based tube current modulation protocol with FBP or IR by random number table. 99 patient with FBP (Group 2:
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Fig. 1. Flow chart of patient inclusion.
351
CCTA: coronary CT angiography; FBP: filter back projection; IR: iterative reconstruction.
48 male, 51 female, age from 37 to 82 with mean of 59.0 years, BMI from 17.9 to 33.3 with mean of 25.5 kg/m2 ) and 93 patent with IR (Group 3: 53 male, 41 female, age from 32 to 83 with mean of 59.6 years, BMI from 16.4 to 35.8 with mean of 25.2 kg/m2 ) were successfully scanned.
2.2.2. Scan protocol and image reconstruction Patients in Group 2 and Group 3 following the same scan parameters as Group 1, except the full tube current being individually calculated with noise-based tube current modulation formulas based on the noise measured from the test bolus image. A flow chart (Fig. 1) illustrates the 3 different acquisition protocols used in our study. Since not all tube current settings were available, we choose the tube current closest to the suggested value.
2.2.3. Qualitative image quality CCTA image data were transferred to an offline workstation (AW 4.5 Advantage, GE Healthcare, Milwaukee, Wisconsin, USA) and independently assessed by two cardiovascular radiologists (Pengyu Li and Jiabin Liu, each with 6 years of experience) who were blinded to technical and clinical information. Curved multiplanar reformations were assessed for each of the four main arteries (left main, left anterior descending, left circumflex and right coronary artery) of each patient. Signal intensity at the aortic root was chosen as
the new window center whereas the window width was arbitrarily defined between +1000 HU and +2000 HU. According to the diagnostic acceptability and the Likert score [9], a 5-point score (Fig. 2) assigned on the basis of the worst scored artery was used: 1 = impaired image quality limited by excessive lumen noise and poor vessel wall definition or poor lumen attenuation, unacceptable for diagnosis; 2 = reduced image quality with massive lumen noise and poor vessel wall definition or low lumen attenuation, diagnosis acceptable only under limited conditions; 3 = adequate image quality with moderate lumen noise and minimal limitation of vessel wall definition, acceptable for diagnosis; 4 = good image quality with minimal lumen noise and well maintained vessel wall definition, fully acceptable for diagnosis; and 5 = excellent image quality with clear vessel wall definition and limited perceptible lumen noise, fully acceptable for diagnosis. Any artery was considered acceptable for diagnosis if image quality was adequate for quantification of luminal stenosis with a minimum diameter of 1.5 mm. Segments with stent graft and extensive calcifications were excluded from analysis. The readers were instructed to ignore motion artifact and stair-step artifact. The disagreement between the two readers was settled by a consensus reading. 2.2.4. Quantitative image quality The measurement of signal intensity (MeanCCTA ) and image noise (SDCCTA ) were performed by placing a circular ROI of about
352
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Table 1 Noise-based tube current modulation formulas regarding SDTB and mACCTA. Group
SDTB (HU)
mACCTA (mA)
Fixed tube current with FBP (Group 1) Noise-based tube current with FBP (Group 2) Noise-based tube current with IR (Group 3)
19.48 (17.07–22.72)
600
P value
19.00 (16.88–21.71)
600 × [(9.550 + 1.177SDTB )/35]2
<0.0001
1.006–1.347
0.808
19.07 (16.40–22.21)
600 × [(8.365 + 0.856SDTB )/35]2
<0.0001
0.722–0.989
0.786
–
95% CI –
r value –
Note: SDTB are median with interquartile range (IQR). SDTB : timing bolus noise; mACCTA : tube current of CCTA; 95%CI: 95% confidence interval for confidence.
100 mm2 in the center of aortic root at the level of the left main artery. MeanCCTA was derived from the CT attenuation values on axial source CCTA images. SDCCTA was defined as the standard deviations of CT attenuation values on axial source CCTA images. Signal-to-noise ratio was calculated as MeanCCTA divided by SDCCTA . All Quantitative image quality was measured by one experienced reader (Lizhen Cao, 9 years of experience). 2.2.5. Radiation exposure Radiation exposure for CCTA was determined by volume CT dose index (CTDIvol), dose-length product (DLP), and effective dose. The DLP was converted to mSv by multiplying it by a chest-specific conversion coefficient of 0.014 mSv/(mGy cm) [11]. 2.2.6. Statistical analysis Statistical analysis was performed using SPSS software (18.0, SPSS Inc., Chicago, III). A P value of less than 0.05 was considered statistically significant difference. Quantitative variables were expressed as mean ± SD and median with interquartile range (IQR) as appropriate; categorical variables were expressed as counts (or proportions in percent). Ordinal variables (i.e., image quality) were also given in means ± SD. The differences between the three groups were assessed with one-way ANOVA for normally distributed variables, and Kruskal–Wallis test for non-normally distributed quantitative and
ordinal variables. For categorical variables, we use chi-square test. For subgroup analysis, a Bonferroni adjustment was made for multiple comparisons. Interobserver agreement for qualitative image quality was estimate using kappa statistic.
3. Results 3.1. Noise-based tube current reduction method Simple linear regression analysis using image noise measurements in Group 1 resulted in equation relating SDCCTA to SDTB . The inverse square relationship between image noise and radiation dose resulted in equation relating SDCCTA to mACCTA . Then, combining the two equations above, noise-based tube current modulation formulas for Group2 and Group3 were established (Table 1). With the noise-based tube current reduction method, the required mACCTA to obtain desired image noise level for CCTA image can be obtained from SDTB on test bolus image. We chose 35 HU as the desired noise level (SD0 ) for CCTA. 3.2. Patient and scan characteristics A total of 294 patients were successfully scanned without adverse events. Comparisons of patient and scan characteristics among three groups are listed in Table 2. No significant differences were observed in terms of patient and scan characteristics, except for tube current, which was lowest for Group 3 (median, 299 mA; IQR, 244–364 mA) and highest for Group 1 (600 mA).
3.3. Qualitative and quantitative image quality
Fig. 2. Quantitative image quality measurement. Region of interest (circle) in the center of aortic root at the level of the left main coronary artery illustrating measurements of signal intensity (434 HU) and image noise (35 HU), as perfromed in every patient.
Interobserver agreement for image quality between the 2 readers was good ( = 0.76). The mean image quality score was 3.38 ± 0.73 for Group 1, which was significantly higher than Group 2 (3.11 ± 0.35, p < 0.05) and Group 3 (3.10 ± 0.33, p < 0.05) (Table 3). However, both Group 2 and Group 3 were scored adequate image quality and similar image quality score was found between the two groups (Fig. 3). Notably, the standard deviation of the image quality score in Group 2 and Group 3 was smaller than that of the Group 1, reflecting more consistent image quality obtained in Group 2 and Group 3. Quantitative analysis demonstrated that no significant difference in signal intensity was observed among three groups, while image noise was significantly lower for Group 1 compared with Group 2 and Group 3. Median image noise was precisely maintained at 35.00 for both Group 2 and Group 3. The IQR of the image noise in Group 2 and Group 3 was much smaller than that of Group 1, reflecting more consistent image noise obtained in Group 2 and Group 3. Signal-to-noise ratio for Group 1 (13.1 ± 3.8) was higher compared with Group 2 (11.9 ± 2.2; p = 0.013) and Group 3 (12.21; p = 0.036). However, Group 3 and Group 2 were comparable regarding signalto-noise ratio (p = 1.000).
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
353
Table 2 Patient and scan characteristics. Characteristics
Fixed tube current with FBP (Group 1)
Noise-based tube current with FBP (Group 2)
Noise-based tube current with IR (Group 3)
n Age, years Male sex Heart ratea , beat/min Height, m Weight, kg Body mass index, kg/m2 Tube current, mA Scan length, mm Pitch Phase (70–80%)
102 57.7 ± 9.5 43.1 (44/102) 56.8 ± 5.6 1.65 ± 0.08 69.8 ± 12.1 25.5 ± 3.4 600 125.1 ± 13.2 0.22 ± 0.01 72.5 (74/102)
99 59.0 ± 9.0 48.5 (48/99) 56.3 ± 5.4 1.65 ± 0.08 69.2 ± 10.1 25.5 ± 2.9 484 (425–604) 127.2 ± 10.9 0.22 ± 0.02 73.7 (73/99)
93 59.6 ± 9.0 55.9 (52/93) 57.2 ± 6.2 1.66 ± 0.07 69.2 ± 10.3 25.2 ± 3.3 299 (244–364) 128.8 ± 12.1 0.22 ± 0.02 67.7 (63/93)
P value
0.248 0.203 0.562 0.624 0.902 0.704 <0.0001 0.113 0.486 0.625
Data are mean ± standard deviation (SD), median with interquartile range (IQR) and percentage with raw data in parentheses. a Heart rate is mean heart rate during scanning.
3.4. Radiation dose estimates CTDIvol was lowest for Group 3 (median, 30.0 mGy; IQR, 24.8–33.7 mGy), followed by Group 2 (median, 43.4 mGy; IQR, 37.3–53.3 mGy) and the by Group 1 (median, 51.0 mGy; IQR, 47.6–58.8 mGy; all p < 0.05). Similarly, the DLP was significant lower for Group 2 (median, 704 mGy cm; IQR, 627–828 mGy cm) compared to Group 1 (median, 822 mGy cm; IQR, 764–924 mGy cm), and further reduced by Group 3 (median, 487 mGy cm; IQR, 407–547 mGy cm; all p < 0.05) compared to Group 2. This corresponds to a relative 15% and 41% reduction in effective dose for Group 2 (median, 9.9 mSv; IQR, 8.8–11.6 mSv) and Group 3 (median, 6.8 mSv; IQR, 5.7–7.7 mSv) compared with Group 1 (median, 11.5 mSv; IQR, 10.7–12.9 mSv; both p < 0.05). That is, the effective dose can be further reduced by a relative 31% for Group 3 compared to Group 2 (p < 0.05).
4. Discussion In this prospective single center study, we established the noisebased tube current reduction methods with and without IR, and then compared protocols with and without noise-based tube current and that with and without IR concerning image quality and radiation exposure. We demonstrate that the use of the noisebased tube current modulation could reduce the effective dose by 14%, compared with fixed tube current protocol while maintaining clinically acceptable images with consistent image noise (35 HU) across the entire patient population, and the effective dose could be further reduced by 31% with IR algorithm. Radiation exposure is a major concern for CCTA especially in case of repeated examinations or in particular subgroups of patients, such as children and young women [1]. A worldwide radiation dose survey indicated that CCTA scan is associated with a median dose of 12 mSv [12]. Accordingly, the median radiation dose of 11.5 mSv
(IQR, 10.7–12.9 mSv) in the fixed tube current protocol of our study compares well with the previous study. Fixed tube current protocol, generally designed to obtain sufficient quality for all patients, leads to unnecessary radiation exposure in the majority of patients, in particular slim. Therefore, modulation of tube current individually is essential for radiation dose optimization. Tube current modulation according to body mass index or weight has been recognized as the standard method and yielded relative good result [13]. However, due to the difference in chest attenuation between men and women, and differences in fat distribution among individuals, the accuracy of this approach is limited [14,15]. In addition, some cardiac and pericardial diseases (e.g. hypertrophic cardiomyopathy, dilated cardiomyopathy and pericardial effusion) can also cause the inaccuracy of the approach above. Noise-based tube current reduction method used in our study modulates the tube current individually according to test bolus image. The noise measured on test bolus image directly reflects the chest attenuation in each patient. Therefore, more accurate tube current modulation and radiation dose optimization can be achieved for subsequent CCTA [8]. Our results show that tube current was significant lower for protocol with noise-based tube current than protocol with fixed tube current using the same FBP algorithm, leading to a significant reduction of effective dose by 14%. Thought radiation dose optimization with noise-based tube current modulation leads to significant dose reduction in our study, it does not necessary mean dose reduction for all patients; for some extreme large patients more radiation dose was required to produce the desired image quality. The relative reduction of effective dose in our study (14%) was lower than that of Qi et al. (28%) [8], which may be due to the statistical method used to establish the noise-based tube current formulas. Specifically, simple linear regression in this study was performed to assess the linear relationship of image noise of CCTA as a function of image noise of test bolus. This was different from the correlation analysis adopted by the study from Qi et al,
Table 3 Results on image quality and radiation exposure. Characteristics
Fixed tube current with FBP (Group 1)
Noise-based tube current with FBP (Group 2)
Noise-based tube current with IR (Group 3)
P value
Image quality score Signal intensity, HU Image noise, HU Signal-to-noise ratio CTDIvol, mGy DLP, mGy cm Effective dose, mSv
3.38 ± 0.73 416 ± 70 32.48 (28.41–36.49) 13.1 ± 3.8 51.0 (47.6–58.8) 822 (764–924) 11.5 (10.7–12.9)
3.11 ± 0.35 413 ± 70 35.00 (34.17–35.08) 11.9 ± 2.2 43.4 (37.3–53.3) 704 (627–828) 9.9 (8.8–11.6)
3.10 ± 0.33 417 ± 67 35.00 (34.34–35.03) 12.1 ± 2.1 30.0 (24.8–33.7) 487 (407–547) 6.8 (5.7–7.7)
<0.0001 0.884 0.001 0.007 <0.0001 <0.0001 <0.0001
Note: CTDIvol: volume CT dose index; DLP: dose-length product. Data are mean ± standard deviation (SD), median with interquartile range (IQR) in parentheses.
354
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
Fig. 3. Representative CCTA demonstrating image quality with different scan protocols. 60 year old woman (body mass index, 25.68 kg/m2 ) (a–c) and 60 year old man (body mass index, 22.03 kg/m2 ) (d–f) both with chest pain. Curved multiplanar reformations of left descending arteries (a, d), left circumflex arteries (b, e), and right coronary arteries (c, f) of two patients demonstrating the same noise level (35 HU) and image quality (score 3) obtained by the noise-based tube current method with (374 mA, a–c) and without iterative reconstruction (189 mA, d–f).
where the purpose was to examine the strength and direction of the relationship between the two variables [16]. The noise-based tube current reduction method has been proposed to determine the required tube current for a desired image noise individually. However, the desired noise level has not been scientifically supported in CCTA and should be determined on the basis of the study purpose (e.g. ruling out significant stenosis, quantification of stenosis, or plaque characteristic analysis) [14]. In order to keep radiation dose as low as reasonably achievable
(ALARA), we chose 35 HU as the desired noise level for CCTA image from the long term experience of our institute, which was higher than previous studies of 27–28 HU [8,17], since it was supposed enough for quantification of stenosis. In consistent with the desired image noise level, the median image noise in this study was precisely maintained at 35 HU with a small amount of variability [(IQR, 34.17–35.08 HU) with FBP and (IQR, 34.34–35.03 HU) with IR], reflecting more consistent image noise achieved using noise-based tube current method. Though image quality score was
J. Shen et al. / European Journal of Radiology 82 (2013) 349–355
significantly lower with noise-based tube current method in this study, it was still adequate for diagnosis. More importantly, the consistent image noise achieved using noise-based tube current method leads to more uniform qualitative image quality than that with fixed tube current protocol. Despite the development of several technological innovations for CCTA within the past few years, dose reduction remains limited by the use of FBP algorithm, which produces a significant increase of image noise in the case of excessive reduction of radiation dose [6]. By iterative comparison of each synthesized forward projection to the actual measurements and modeling system statistics (photo statistics and electronic noise), IR has the potential to selectively reduce image noise, which may permit preserved image quality with reduced tube current, thereby permitting lower radiation dose [7]. In consistence with the previous study, IR was not associated with decrease in signal and SNR. In our study, noise-based tube current method with IR algorithm saved 41% effective dose compared with fixed tube current protocol using FBP algorithm. Though higher effective does reduction (44%) was observed between the two algorithms in the ERASIR study [18]. The difference in relative dose savings may be due to the other dose reduction techniques (i.e., sequential scan mode, 100 kV tube voltage) employed in the ERASIR study. Therefore, after adjustment for these scan settings, IR was only associated with a 27% reduction in effective dose in comparison to FBP. In addition, though no significant difference was observed in image noise in the ERASIR study, it was not intended to maintain at the same level for each patient and tube current was merely controlled based on experience of each study site. On contrast, our study made use of the noise-based method to modulate the tube current individually and maintain the median noise level precisely at 35 HU with small IQR, which effectively eliminating difference in chest attenuation and leading to the consistent image quality for each patient. Our study had several limitations. First, the radiation dose associated with the scout scan, calcium score scan, and test bolus scan was not assessed in this study. However, compared with CCTA, the radiation dose of these pre-control scans represents only a small fraction of the total dose. Second, the retrospectively gated CCTA was used in addition to assess left ventricular function. Compared to retrospectively gating technique, studies with prospective gating have reported a 69% effective dose reduction [19]. Noise-based tube current reduction method can also be applied for prospective gated CCTA. Future studies are warranted to investigate this method in prospective gated CCTA. Third, the IR algorithm used in our study represents a modified and limited form of IR. More complete IR can further improve image quality with substantial decreased image noise, allowing further reduction in radiation dose [20]. However, high computational cost and long reconstruction times remain a barrier for the complete IR in daily clinical application. Forth, radiation dose varied with the square of tube voltage in the setting of a constant tube current. Therefore, reducing tube voltage has a greater effect on the reduction of radiation exposure than reducing tube current. Radiation dose was relatively high in our study because the tube voltage was fixed at 120 kVp and only tube current was modulated. Future studies are warranted to integrate low tube voltage (100 kVp or 80 kVp) into the noise-based tube current modulation method. 5. Conclusions We firstly integrate IR algorithm into noise-based tube current method to optimize radiation dose in CCTA. Therefore, image noise could be precisely maintained at the desired level individually and consistent image quality could be obtained across the entire patient
355
population. At the same time, a 41% reduction in effective dose is achieved compared with a fixed tube current protocol. Conflict of interest No conflict of interest exits in the submission of this manuscript. Acknowledgment The authors would like to thank Dr Jianying Li, Huizhi Cao and Shichen Liu for technique suggestions and language revision. Our work was supported by the National Natural Science Foundation (30970821) and National Basic Research Program 973 (2010CB732600) of China. References [1] Hausleiter J, Martinoff S, Hadamitzky M, et al. Image quality and radiation exposure with a low tube voltage protocol for coronary CT angiography results of the PROTECTION II Trial. JACC: Cardiovascular Imaging 2010;3(11): 1113–23. [2] Jakobs TF, Becker CR, Ohnesorge B, et al. Multislice helical CT of the heart with retrospective ECG gating: reduction of radiation exposure by ECG-controlled tube current modulation. European Radiology 2002;12(5):1081–6. [3] Earls JP, Berman EL, Urban BA, et al. Prospectively gated transverse coronary CT angiography versus retrospectively gated helical technique: improved image quality and reduced radiation dose. Radiology 2008;246(3):742–53. [4] Alkadhi H, Stolzmann P, Desbiolles L, et al. Low-dose, 128-slice, dual-source CT coronary angiography: accuracy and radiation dose of the high-pitch and the step-and-shoot mode. Heart 2010;96(12):933–8. [5] Matthew JW, Mark EO, Milind YD, et al. New radiation dose saving technologies for 256-slice cardiac computed tomography angiography. International Journal of Cardiovascular Imaging 2009;25(Suppl. 2):189–99. [6] Gervaise A, Osemont B, Lecocq S, et al. CT image quality improvement using adaptive iterative dose reduction with wide-volume acquisition on 320detector CT. European Radiology 2012;22(2):295–301. [7] Hara AK, Paden RG, Silva AC, et al. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. American Journal of Roentgenology 2009;193(3):764–71. [8] Qi W, Li J, Du X. Method for automatic tube current selection for obtaining a consistent image quality and dose optimization in a cardiac multidetector CT. Korean Journal of Radiology 2009;10(6):568–74. [9] Leipsic J, Labounty TM, Heilbron B, et al. Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography. American Journal of Roentgenology 2010;195(3):649–54. [10] Boedeker KL, Mcnitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part II. Noise power spectra and signal to noise. Physics in Medicine and Biology 2007;52(14):4047–61. [11] American Association of Medical Physicists (AAPM) Task Group 23. CT dosimetry. The measurement, reporting and management of radiation dose in CT. Available at: http://www.aapm.org/pubs/reports (accessed 01.08.10). [12] Hausleiter J, Meyer T, Hermann F, et al. Estimated radiation dose associated with cardiac CT angiography. JAMA 2009;301(5):500–7. [13] Tatsugami F, Husmann L, Herzog BA, et al. Evaluation of a body mass index-adapted protocol for low-dose 64-MDCT coronary angiography with prospective ECG triggering. American Journal of Roentgenology 2009;192(3):635–8. [14] Hur G, Hong SW, Kim SY, et al. Uniform image quality achieved by tube current modulation using SD of attenuation in coronary CT angiography. American Journal of Roentgenology 2007;189(1):188–96. [15] Paul JF, Abada HT. Strategies for reduction of radiation dose in cardiac multislice CT. European Radiology 2007;17(8):2028–37. [16] Zou KH, Tuncali K, Silverman SG. Correlation and simple linear regression. Radiology 2003;227(3):617–28. [17] Gao J, Li J, Earls J, et al. Individualized tube current selection for 64-row cardiac CTA based on analysis of the scout view. European Journal of Radiology 2011;79(2):266–71. [18] Leipsic J, Labounty TM, Heilbron B, et al. Estimated radiation dose reduction using adaptive statistical iterative reconstruction in coronary CT angiography: the ERASIR study. American Journal of Roentgenology 2010;195(3): 655–60. [19] Hausleiter J, Meyer TS, Martuscelli E, et al. Image quality and radiation exposure with prospectively ECG-triggered axial scanning for coronary CT angiography: the multicenter, multivendor. randomized PROTECTION-III study. JACC: Cardiovascular Imaging 2012;5(5):484–93. [20] Scheffel H, Stolzmann P, Schlett CL, et al. Coronary artery plaques: cardiac CT with model-based and adaptive-statistical iterative reconstruction technique. European Journal of Radiology 2011;81(3):363–9.