Image quality of ct angiography using model-based iterative reconstruction in infants with congenital heart disease: Comparison with filtered back projection and hybrid iterative reconstruction

Accepted Manuscript Title: Image quality of ct angiography using model-based iterative reconstruction in infants with congenital heart disease: Comparison with filtered back projection and hybrid iterative reconstruction Authors: Qianjun Jia MD Jian Zhuang MD, PHD Jun Jiang MD Jiahua Li MD Meiping Huang MD, PHD Dr. Changhong Liang MD, PHD PII: DOI: Reference:

S0720-048X(16)30322-9 http://dx.doi.org/doi:10.1016/j.ejrad.2016.10.017 EURR 7601

To appear in:

European Journal of Radiology

Received date: Revised date: Accepted date:

10-5-2016 2-8-2016 15-10-2016

Please cite this article as: Jia Qianjun, Zhuang Jian, Jiang Jun, Li Jiahua, Huang Meiping, Liang Changhong.Image quality of ct angiography using model-based iterative reconstruction in infants with congenital heart disease: Comparison with filtered back projection and hybrid iterative reconstruction.European Journal of Radiology http://dx.doi.org/10.1016/j.ejrad.2016.10.017 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

Title page Image Quality of CT Angiography Using Model-based Iterative Reconstruction in Infants with Congenital Heart Disease: Comparison with Filtered Back Projection and Hybrid Iterative Reconstruction

Authors and Affiliations: Qianjun Jia, MD: Southern Medical University, Guangzhou, Guangdong, China. Department of Radiology, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Department of Catheterization Lab, Guangdong Cardiovascular Institute, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Tel: +86 15802036497 Fax: +86-020-83870125 E-mail: [email protected]

Jian Zhuang MD, PHD: Department of Cardiac Surgery, Guangdong Cardiovascular Institute, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Tel: +86-020-83870125

2

Fax: +86-020-83870125 E-mail: [email protected]

Jun Jiang MD: Department of Radiology, Shenzhen Second People's Hospital, Shenzhen, Guangdong, China. Tel: +86-13689578352 Fax: +86-020-83870125 E-mail: [email protected]

Jiahua Li, MD: Department of Catheterization Lab, Guangdong Cardiovascular Institute, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Tel: +86-020-83870125 Fax: +86-020-83870125 E-mail: [email protected]

Meiping Huang, MD, PHD: Department of Catheterization Lab, Guangdong Cardiovascular Institute, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Southern Medical University, Guangzhou, Guangdong, China.

3

Tel: +86-020-83870125 or +86-13719270493 Fax: +86-020-83870125 E-mail: [email protected]

Changhong Liang MD, PHD Department of Radiology, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Southern Medical University, Guangzhou, Guangdong, China. Telephone: +86-020-83870125 or +86-13902215278 Fax: +86-020-83870125 E-mail: [email protected]

Mailing Address for Correspondence and Reprints: Dr. Changhong Liang, Department of Radiology, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Telephone: +86-020-83870125 or +86-13902215278 Fax: +86-020-83870125 E-mail: [email protected] Dr. Meiping Huang, Department of Catheterization Lab, Guangdong Cardiovascular Institute, Guangdong Provincial Key Laboratory of South China Structural Heart Disease, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, Guangdong, China. Tel: +86-020-83870125 or +86-13719270493

4

Fax: +86-020-83870125 E-mail: [email protected]

5

Image Quality of CT Angiography Using Model-based Iterative Reconstruction in Infants with Congenital Heart Disease: A Comparison with Filtered Back Projection and Hybrid Iterative Reconstruction

Abstract Purpose: To compare the image quality, rate of coronary artery visualization and diagnostic accuracy of 256-slice multi-detector computed tomography angiography (CTA) with prospective electrocardiographic (ECG) triggering at a tube voltage of 80 kVp between 3 reconstruction algorithms (filtered back projection (FBP), hybrid iterative reconstruction (iDose4) and iterative model reconstruction (IMR)) in infants with congenital heart disease (CHD). Methods: Fifty-one infants with CHD who underwent cardiac CTA in our institution between December 2014 and March 2015 were included. The effective radiation doses were calculated. Imaging data were reconstructed using the FBP, iDose4 and IMR algorithms. Parameters of objective image quality (noise, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR)); subjective image quality (overall image quality, image noise and margin sharpness); coronary artery visibility; and diagnostic accuracy for the three algorithms were measured and compared. Results: The mean effective radiation dose was 0.61±0.32 mSv. Compared to FBP and iDose4, IMR yielded significantly lower noise (P<0.01), higher SNR and CNR values (P<0.01), and a greater subjective image quality score (P<0.01). The total number of coronary segments visualized was significantly higher for both iDose4 and IMR than for FBP (P=0.002 and P=0.025, respectively), but there was no significant difference in this parameter between iDose4 and IMR (P=0.397). There was

6

no significant difference in the diagnostic accuracy between the FBP, iDose4 and IMR algorithms (2=0.343, P=0.842). Conclusions: For infants with CHD undergoing cardiac CTA, the IMR reconstruction algorithm provided significantly increased objective and subjective image quality compared with the FBP and iDose4 algorithms. However, IMR did not improve the diagnostic accuracy or coronary artery visualization compared with iDose4. Key words: Congenital heart disease; Multidetector CT; Hybrid iterative reconstruction; Iterative model reconstruction Abbreviations:

CHD=Congenital heart disease

FBP=Filtered back projection

IR= Iterative Reconstruction

HIR= Hybrid iterative reconstruction

IMR= Iterative model reconstruction

SNR=Signal-to-noise ratio

CNR=Contrast-to-noise ratio

CCA=Conventional cardiac angiography

7

ROI=Region of interest

8

Introduction Recent and rapid advances in multi-detector computed tomography (MDCT) technology that have improved spatial and temporal resolution [1, 2] , thus enabling cardiac CT examinations in paediatric patients with congenital heart disease (CHD). However, the radiation exposure delivered by CT is a concern, particularly for children and neonates, who are more radiosensitive than adults and who have an increased lifetime risk of radiation-induced carcinogenesis [3, 4]. Various strategies have been described to reduce the radiation exposure, including low-tube voltage [5], tube-current modulation [6] and prospective electrocardiogram (ECG) triggering [7]. However, most techniques intended to reduce the radiation dose, particularly techniques using decreased tube voltage or tube current, inevitably increase the image noise due to a combination of reduced photon flux and energy [8]. The filtered back projection (FBP) algorithm, an analytical approach to image reconstruction, has long been the standard method for reducing the radiation dose. An iterative reconstruction (IR) algorithm was recently developed to reduce the quantum noise associated with the FBP reconstruction algorithm. Previous studies have shown that a hybrid IR (HIR) algorithm increased image quality and compensated for image noise more effectively than can the FBP reconstruction algorithm [9]. Iterative model reconstruction (IMR) is a recently developed, knowledge-based IR algorithm based on raw data. In contrast to its predecessors such as iDose (Philips), the IMR algorithm, a real IR technique that does not blend with FBP and is more advanced and intricate, makes multiple comparisons between the reconstructed and measured data in the raw data domain and iteratively corrects the images [10-14].

9

However, the applicability of an 80-kVp tube voltage combined with the IMR algorithm for 256-slice multi-detector cardiac CTA in the axial plane using prospective ECG triggering to evaluate infants remains uncertain. Therefore, we conducted this study to compare the image quality and diagnostic accuracy of the IMR, iDose4 and FBP reconstruction algorithms in infants with CHD. Our purpose was to evaluate the performance of IMR and to generate insights into its clinical applications.

Materials and Methods Study population This prospective study was approved by the institutional review board. The possible adverse effects of the contrast medium injection and radiation exposure were explained to the infants’ legal guardians, and written informed consent was obtained for all patients. We enrolled 51 consecutive infants (<1 year old) with CHD referred for cardiac CT in our institution between December 2014 and March 2015. The exclusion criteria were an allergy to the contrast medium or renal insufficiency. All observed anomalies were confirmed via surgery and/or conventional cardiac angiography (CCA). Multi-detector CTA protocol Images were obtained using a 256-slice MDCT scanner (Brilliance iCT; Philips Healthcare, Cleveland, OH, USA). Details regarding the cardiac CTA protocol that was used have been reported previously [15] .

10

CT image reconstruction Images from the most quiescent phases of the scans were transferred to an external workstation (Cardiac Viewer, Extended Brilliance Workspace (version 4.0); Philips Healthcare, Cleveland, OH, USA) for interpretation. FBP, HIR (iDose4; Philips Healthcare), and knowledge-based IMR (IMR; Philips Healthcare) algorithms were used. For FBP and HIR, images were reconstructed with a kernel routinely used in clinical practice. We applied the moderate-level HIR algorithm that is routinely used at our institute (iDose4; Level 4, with a noise reduction factor of 0.29). IMR has two groups of reconstruction settings, routine and sharp. Because we were using cardiac scans, a setting specific to the cardiac anatomy, referred to as "cardiac routine", was used for the IMR algorithm-based image reconstructions. Each setting has three levels of noise reduction (Level 1 (least noise reduction), Level 2 and Level 3 (greatest noise reduction)). Because our target in this study was moderate noise reduction, we used Level 2. CT radiation dose estimates The volume CT dose index (CTDIvol) and dose-length product (DLP) values during the scans were recorded. The estimated effective dose (ED) was derived from the DLP and the conversion coefficient k (ED=DLP×k; k=0.039 for infants up to 4 months of age, and k=0.026 for infants between 4 months and 1 year of age) [16]. Objective image evaluation To evaluate the objective image quality, regions of interest (ROIs) that were as large as the diameter of the lumen were drawn in the

11

ascending aorta, descending aorta, main pulmonary artery, right atrium, right ventricle, left atrium and left ventricle. The attenuation value was measured on the axial images. The attenuation, noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) in 7 equivalent positions were measured by 1 observer who was blinded to the subjective image quality results. Image noise was defined as the standard deviation of the attenuation value, and the SNR and CNR were calculated as follows: SNR=mean_heart or vessel/SD_muscle, and CNR=(mean_heart or vessel-mean_muscle)/SD_muscle, where mean_heart or vessel is the mean CT value of the heart or vessels, mean_muscle is the mean CT value of the chest muscle, and SD is the standard deviation of the ROI. Subjective image quality analysis Images reconstructed using the FBP, iDose4 and IMR algorithms were reviewed in a random order. Two radiologists, each with more than 10 years of experience in cardiac imaging, independently interpreted the image quality of the axial, multiplanar reformation (MPR), maximum intensity projection and volume rendering images. Any disagreements between the two observers were resolved by consensus. Overall CT image quality, with the exception of images of the coronary arteries, was evaluated using the following five-grade scoring system: Grade 1=no useful information obtained; Grade2=poor image quality or anatomical detail, resulting in incomplete demonstration of anatomical structures; Grade 3=fair anatomical clarity such that the clinically significant anatomical relationships required clinically could be defined with confidence; Grade 4=good anatomical clarity, such that all structures were clearly interpretable; and Grade 5=excellent anatomical clarity, representing excellent image quality. A five-point subjective scale was used for noise assessment: 1=severe and unacceptable noise; 2=

12

noise interfering with the depiction of the cardiovascular structures; 3=moderate noise not interfering with the depiction of the cardiovascular structures; 4=below average noise; and 5=minimal or no noise. Margin sharpness was assessed as follows: 1=blurry and unacceptable; 2=below average; 3=average; 4=above average (between scores 3 and 5); and 5=sharpest [17]. Coronary artery evaluation For each patient, the coronary artery was divided into 7 segments: the left main (LM) coronary artery, the proximal and distal segments of the left anterior descending (LAD) coronary artery, the proximal and distal segments of the left circumflex coronary artery (LCX), and the proximal and distal segments of the right coronary artery (RCA). Based on the 17-segment model presented in the Society of Cardiovascular Computed Tomography guidelines [18], we considered segments 1 and 2; segments 6, and 7; and segments 11, 12, and 17 as the proximal RCA, proximal LAD, and proximal LCX, respectively. Subjective rating scores of the coronary arteries were obtained using the following 5-point scale: 1=non-diagnostic image quality (high visible effects of noise and severe contour and contrast); 2=poor image quality (moderate effects of noise and poor contour and contrast); 3=moderate image quality (mild visible effects of noise and moderate contour and contrast); 4=good image quality (minimal visible effects of noise and good contour and contrast); and 5=excellent image quality (no visible effects of noise and excellent contour and contrast). An artery with an image score of 3, 4, or 5 was considered detectable, while an artery with an image score of 1 or 2 was considered non-detectable. Diagnostic performance analysis

13

Two radiologists who were not involved in the image quality assessment and who were blinded to the results of the surgical and/or CCA findings evaluated all images in consensus. The FBP, iDose4 and IMR image series were presented in random order. Using the surgical and/or CCA findings as the reference standard, the diagnostic accuracy was calculated and compared among the FBP, iDose4 and IMR algorithms.

Statistical analysis Statistical analysis was performed using SPSS 16.0 software (SPSS; Chicago, IL, USA). All numeric values were reported as the means±standard deviation. For normally and non-normally distributed data, differences in the mean values of the objective image quality parameters between the three reconstruction algorithms were determined using the Tukey-Kramer and Steel-Dwass tests, respectively. The inter-observer agreement of the 2 observers in the visual evaluation of the subjective image quality (image noise, margin sharpness and overall image quality) and coronary artery visibility was measured with the kappa statistic. A kappa value greater than 0.81 corresponded to excellent inter-observer agreement, and a kappa value of 0.61 to 0.80 corresponded to good inter-observer agreement. The subjective image quality scores assigned to the three groups of images were compared using the Friedman test. If there were significant differences between the groups, pairwise comparisons were performed using the Steel-Dwass test. Comparative analyses of the diagnostic performance and the rate of the coronary artery visualization between the FBP, iDose4 and IMR algorithms were performed using a non-parametric chi-square test. A differences of P<0.05 was considered statistically significant.

14

Results Patient population and radiation dose

All 51 patients (29 male, 22 female) underwent successful CTA, and no complications occurred as a result of the scans. The mean age was 4.36±4.09 months (range, 3 days to 12 months), and the mean weight was 5.64±2.10 kg (range, 2.20-11.20 kg). The patient demographics and radiation doses are shown in Table 1. The mean CTDIvol was 1.59±0.49 mGy, and the mean DLP was 18.27±7.57 mGy cm, which corresponded to a mean estimated effective dose of 0.61±0.32 mSv. Objective image quality The mean density and image noise results are shown in Table 2. The objective image noise in the IMR images of the aorta, pulmonary trunk, and heart chambers was significantly lower than that in the iDose4 and FBP images (all P<0.01). However, no significant difference in the attenuation of the four heart chambers, the ascending aorta, the descending aorta, or the pulmonary trunk was observed between the 3 groups (all P>0.05) (Table 2). The CNR and SNR values were much higher for IMR than for the other 2 reconstruction methods (all P>0.05) (Table 2). Compared to iDose4 and FBP, IMR showed a 139.91-278.86% increase in CNR and a 21.21-154.03% decrease in noise. Subjective image quality The results of our subjective image quality assessment are shown in Table 3. There was excellent inter-observer agreement for image noise,

15

margin sharpness and overall image quality (kappa=0.81, 0.88 and 0.86, respectively). The visual scores for the subjective image quality parameters (image noise, margin sharpness and overall image quality) were significantly higher for IMR than for FBP and iDose4 (all P<0.01). A representative case is shown in Figure 1. Coronary artery evaluation The number of detected coronary artery segments (image score of the coronary artery segments ≥3) is shown in Table 4. There was good-to-excellent inter-observer agreement for the visibility of the LM, proximal and distal LAD, proximal and distal LCX, and proximal and distal RCA visibility (kappa=0.84, 0.78, 0.72, 0.75, 0.72, 0.77 and 0.73, respectively). The total number of coronary segments visualized was significantly higher for both iDose4 and IMR than for FBP (P=0.002 and P=0.025, respectively), but there was no significant difference in this parameter between iDose4 and IMR (P=0.397). There was no significant difference in visualization of the LM (P=0.862), LAD (P=0.056), LCX (P=0.147) or RCA (P=0.163) between FBP, iDose4 and IMR. IMR showed slightly fewer total segments and poorer LAD visualization than iDose4 (those results were statistically insignificant: P=0.397 and P=0.272, respectively). A representative case is shown in Figure 2. Diagnostic accuracy A total of 193 separate cardiovascular anomalies were confirmed based on the surgical and/or CCA results. The details of the cardiovascular abnormalities are shown in Table 5. There was no significant difference in the diagnostic accuracy between the FBP, iDose4 and IMR algorithms (96.89%, 97.41% and 96.37% respectively; 2 = 0.343, P=0.842). According to the FBP and IMR image series, 3 small atrial septal defects

16

(ASDs) and 2 small ventricular septal defects (VSDs) were misdiagnosed as normal. According to the iDose4 image series, 2 ASDs and 2 VSDs were misdiagnosed as normal. The FBP and iDose4 image series failed to identify 1 case of patent ductus arteriosus (PDA), while 2 cases of PDA were missed using the IMR image series (Figure 3).

17

Discussion Our pilot study showed that IMR yielded the highest CNR and SNR values as well as the greatest visualization scores for the assessment of large vessels and heart chambers, but the subjective image quality scores of the coronary vessels did not differ significantly between the algorithms tested. It is possible to reduce the radiation dose by reducing the tube potential (i.e., peak kilo-voltage) [19] or tube current (i.e., milli-ampere). In daily practice, 80 kVp is routinely used for CT examinations of infants and young children [7, 15, 20]. Another option for dose reduction is the use of automatic tube current modulation, which enables the adjustment of the tube current in various planes (x, y, and z or a combination) to the respective attenuation of the body region, with the goal of maintaining constant image quality. In addition to using a tube voltage of 80-kVp tube voltage and adapting the tube current to body weight, the prospective ECG-triggering sequential mode was applied in our study. Our previous study [15] demonstrated that low-dose, prospectively gated axial 256-slice CTA is valuable for the routine clinical evaluation of infants with CHD and provides a comprehensive, three-dimensional evaluation of the cardiac anatomy. A previous study of 256-MDCT coronary angiography image quality and radiation dose between the IR and FBP algorithms showed that a tube current reduction of 50% was the optimum cut-off point for a low tube-current strategy under the same peak kilo-voltage [21]. In the present study, the tube current of a given weight-adjusted radiation dose adjustment was only half of that used in our previous study [15]. The downside of low tube-voltage and low tube-current CT scanning is the increased image noise, which may impair diagnostic confidence. To counterbalance this increase in image noise,

18

IR (iDose4 and IMR) were used, although not in our previous study [15]. Full IR involves both forward and backward reconstruction steps. In forward reconstruction, an assumption based on the characteristics of the CT system is made about the distribution of the attenuation in the scanned area. The assumed distribution of the attenuation is then compared with the corresponding measured attenuation distribution, and a more accurate assumption is made based on this comparison [22]. The forward and backward projections are repeated until they do not change after subsequent iterations or until the maximum number of iterations is reached [22]. Ultimately, a final optimized image is reconstructed. The combination of IR and FBP is known as hybrid IR. The hybrid IR technique used in this study (iDose4; Philips Healthcare) uses multi-frequency noise removal techniques in the projection and image domains to help reduce noise uniformly across the complete frequency range, which results in an image appearance or “look” that is a close representation of conventional FBP reconstructions [21]. In contrast to previous generations of image reconstruction algorithms, the IMR algorithm does not involve blending with FBP images, and it is mathematically more complex and accurate. IMR uses a knowledge-based approach to accurately determine the data and image statistics and the system models, which depict the geometry and physical characteristics of the CT scanner [14]. IMR yields optimal images by iteratively minimizing the difference between the acquired data and their ideal form and takes into account the imaging model [23]. The benefits of IMR for various clinical applications have been shown in previous studies [10, 11, 14, 24]. However, there have been few reports on the clinical application of IMR to imaging of infants with CHD. In our study, the image noise was significantly decreased significantly

19

by 21.21-154.03% using IMR compared to FBP and iDose4. This result is supported by those of other earlier studies [17, 25]. According to Yamada et al [26], substantial streak artefacts in adaptive statistical IR images could lead to an increase in image density; however, in the present study, the mean attenuation values did not differ significantly between the 3 methods, and our observation was consistent with the findings of Sonc et al and Yuki et al [24, 25]. In addition, the IMR algorithm yielded the highest CNR and SNR values as well as the greatest visualization scores for the assessment of cardiac structures and large vessels compared to the FBP and iDose4 algorithms. Yuki et al [24], who compared FBP, iDose4 and IMR in cardiac CT imaging, reported that IMR yielded a CNR that was approximately 107-345% higher than that produced using the other two methods. Their results were similar to ours, although the difference in the CNR between IMR and FBP/iDose 4 was smaller in our study. This discrepancy may be because they studied adult hearts, while our study focused on infants. In addition, we used a low tube voltage of 80 kVp, whereas they used a higher tube voltage of 100 kVp. Previous study has shown that IMR produces better-quality images during coronary CTA using a tube voltage of 80 kVp with a prospective ECG gating technique in adults than does IR or FBP [27]. However, because only few studies have used model-based IR in a paediatric setting [25, 28, 29], certain effects of model-based IR techniques remain unclear, such as the visibility of small structures. Although coronary artery anomalies may make intracardiac repair difficult [30], coronary arteries in infants are occasionally difficult to visualize clearly due to their small configuration [31]. In contrast to our expectations, IMR yielded slightly poorer coronary artery visibility in infants than did iDose4 in infants. The exact cause of this slightly poorer coronary artery image quality using the IMR algorithm remains unknown, although we believe that it is at

20

least partly related to the faster heartbeat of an infant, which is the major challenge to visualization of the coronary artery. The mean heart rate was approximately 138 beats per min (bpm) in the present study of infants, whereas the mean heart rate was 55 bpm in the previous adult-based study of adults [27]. We also believe that the smaller coronary arteries of infants than those of adults contributed to these differences between studies. One disadvantage of model-based IR is the smoother appearance of images because details occasionally have a blurred appearance, particularly in images obtained using higher IR levels. The smoothing effect of IR can potentially result in small structures not being visible [32, 33]. The present study showed that the diagnostic accuracy of the FBP, iDose4 and IMR algorithms did not differ significantly and that IMR did not improve the visibility of small structures (small ASDs, VSDs and PDA) relative to FBP and iDose4. We speculate that this phenomenon can be explained by reasons similar to those noted above.

Limitations There were several limitations to our single-centre study. First, our pilot study included a small number of patients, and our techniques must be rigorously evaluated in large-scale clinical investigations. Second, studies that compare IR algorithms with standard FBP cannot be appropriately blinded because of the obvious reduction of noise in the IR-processed images. Third, the 3 levels of iteration presets of IMR (strength 1-3) associated with its ability to reduce noise were not compared, and a further evaluation of different iterative reconstruction levels for the same protocols should be performed. Finally, an image quality evaluation was not performed with a phantom study. Vascular enhancement affects

21

image quality, and therefore, a phantom study would have provided more accurate results.

Conclusions In conclusion, in cardiac CTA performed on infants with CHD, the IMR algorithm yielded significantly higher objective and subjective image quality than the FBP and iDose4 algorithms at an acceptable radiation dose. However, IMR did not improve either diagnostic accuracy or coronary artery visibility compared to iDose4.

Conflict of Interest Statement: We declare that we have no conflict of interest.

Acknowledgements: This work was supported by the National natural science foundation of China (No.U1401255), National Key Technology R&D Program of China (No. 2011BAI11B22), Guangdong Province science and technology planning project of China (No. 2009B030801257, No. 2013B031800006, No. 2014A020212228), Guangzhou city science and technology planning project of China (No. 1563000374) and Guangdong Province Medical Research Foundation (No. A2013036).

22

References [1] P. Ou, D.S. Celermajer, G. Calcagni, F. Brunelle, D. Bonnet, D. Sidi, Three-dimensional CT scanning: a new diagnostic modality in congenital heart disease, Heart 93(8) (2007) 908-13. [2] C. Young, A.M. Taylor, C.M. Owens, Paediatric cardiac computed tomography: a review of imaging techniques and radiation dose consideration, Eur. Radiol. 21(3) (2011) 518-29. [3] M.S. Pearce, J.A. Salotti, M.P. Little, K. McHugh, C. Lee, K.P. Kim, N.L. Howe, C.M. Ronckers, P. Rajaraman, A.W. Sir Craft, L. Parker, A. Berrington de Gonzalez, Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study, Lancet 380(9840) (2012) 499-505. [4] A.J. Einstein, M.J. Henzlova, S. Rajagopalan, Estimating risk of cancer associated with radiation exposure from 64-slice computed tomography coronary angiography, Jama 298(3) (2007) 317-23. [5] S. Leschka, P. Stolzmann, F.T. Schmid, H. Scheffel, B. Stinn, B. Marincek, H. Alkadhi, S. Wildermuth, Low kilovoltage cardiac dual-source CT: attenuation, noise, and radiation dose, Eur. Radiol. 18(9) (2008) 1809-17. [6] C. Herzog, D.M. Mulvihill, S.A. Nguyen, G. Savino, B. Schmidt, P. Costello, T.J. Vogl, U.J. Schoepf, Pediatric cardiovascular CT angiography: radiation dose reduction using automatic anatomic tube current modulation, AJR Am. J. Roentgenol. 190(5) (2008) 1232-40.

23

[7] P. Nie, X. Wang, Z. Cheng, X. Ji, Y. Duan, J. Chen, Accuracy, image quality and radiation dose comparison of high-pitch spiral and sequential acquisition on 128-slice dual-source CT angiography in children with congenital heart disease, Eur. Radiol. 22(10) (2012) 2057-66. [8] B.K. Han, J. Lindberg, K. Grant, R.S. Schwartz, J.R. Lesser, Accuracy and safety of high pitch computed tomography imaging in young children with complex congenital heart disease, Am. J. Cardiol. 107(10) (2011) 1541-6. [9] D. Utsunomiya, W.G. Weigold, G. Weissman, A.J. Taylor, Effect of hybrid iterative reconstruction technique on quantitative and qualitative image analysis at 256-slice prospective gating cardiac CT, Eur. Radiol. 22(6) (2012) 1287-94. [10] R.D. Khawaja, S. Singh, M. Gilman, A. Sharma, S. Do, S. Pourjabbar, A. Padole, D. Lira, K. Brown, J.A. Shepard, M.K. Kalra, Computed tomography (CT) of the chest at less than 1 mSv: an ongoing prospective clinical trial of chest CT at submillisievert radiation doses with iterative model image reconstruction and iDose4 technique, J. Comput. Assist. Tomogr. 38(4) (2014) 613-9. [11] H. Kim, C.M. Park, Y.S. Song, S.M. Lee, J.M. Goo, Influence of radiation dose and iterative reconstruction algorithms for measurement accuracy and reproducibility of pulmonary nodule volumetry: a phantom study, Eur. J. Radiol. 83(5) (2014) 848-57. [12] R.D. Khawaja, S. Singh, M. Blake, M. Harisinghani, G. Choy, A. Karosmangulu, A. Padole, S. Do, K. Brown, R. Thompson, T. Morton, N. Raihani, T. Koehler, M.K. Kalra, Ultra-low dose abdominal MDCT: using a knowledge-based Iterative Model Reconstruction technique for substantial dose reduction in a prospective clinical study, Eur. J. Radiol. 84(1) (2015) 2-10.

24

[13] J.H. Yoon, J.M. Lee, M.H. Yu, J.H. Baek, J.H. Jeon, B.Y. Hur, A. Dhanantwari, S.Y. Chung, J.K. Han, B.I. Choi, Comparison of iterative model-based reconstruction versus conventional filtered back projection and hybrid iterative reconstruction techniques: lesion conspicuity and influence of body size in anthropomorphic liver phantoms, J. Comput. Assist. Tomogr. 38(6) (2014) 859-68. [14] S. Oda, D. Utsunomiya, Y. Funama, K. Katahira, K. Honda, S. Tokuyasu, M. Vembar, H. Yuki, K. Noda, S. Oshima, Y. Yamashita, A knowledge-based iterative model reconstruction algorithm: can super-low-dose cardiac CT be applicable in clinical settings?, Acad. Radiol. 21(1) (2014) 104-10. [15] M.P. Huang, C.H. Liang, Z.J. Zhao, H. Liu, J.L. Li, J.E. Zhang, Y.H. Cui, L. Yang, Q.S. Liu, T.B. Ivanc, M. Vembar, Evaluation of image quality and radiation dose at prospective ECG-triggered axial 256-slice multi-detector CT in infants with congenital heart disease, Pediatr. Radiol. 41(7) (2011) 858-66. [16] G. Pache, J. Grohmann, S. Bulla, R. Arnold, B. Stiller, C. Schlensak, M. Langer, P. Blanke, Prospective electrocardiography-triggered CT angiography of the great thoracic vessels in infants and toddlers with congenital heart disease: feasibility and image quality, Eur. J. Radiol. 80(3) (2011) e440-5. [17] S. Oda, G. Weissman, M. Vembar, W.G. Weigold, Cardiac CT for planning redo cardiac surgery: effect of knowledge-based iterative model reconstruction on image quality, Eur. Radiol. 25(1) (2015) 58-64. [18] G.L. Raff, A. Abidov, S. Achenbach, D.S. Berman, L.M. Boxt, M.J. Budoff, V. Cheng, T. DeFrance, J.C. Hellinger, R.P. Karlsberg, T.

25

Society of Cardiovascular Computed, SCCT guidelines for the interpretation and reporting of coronary computed tomographic angiography, J. Cardiovasc. Comput. Tomogr. 3(2) (2009) 122-36. [19] B. Bischoff, F. Hein, T. Meyer, M. Hadamitzky, S. Martinoff, A. Schomig, J. Hausleiter, Impact of a reduced tube voltage on CT angiography and radiation dose: results of the PROTECTION I study, JACC Cardiovasc. Imaging 2(8) (2009) 940-6. [20] F. Tricarico, A.M. Hlavacek, U.J. Schoepf, U. Ebersberger, J.W. Nance, Jr., R. Vliegenthart, Y.J. Cho, J.R. Spears, F. Secchi, G. Savino, R. Marano, S.O. Schoenberg, L. Bonomo, P. Apfaltrer, Cardiovascular CT angiography in neonates and children: image quality and potential for radiation dose reduction with iterative image reconstruction techniques, Eur. Radiol. 23(5) (2013) 1306-15. [21] Y. Hou, X. Liu, S. Xv, W. Guo, Q. Guo, Comparisons of image quality and radiation dose between iterative reconstruction and filtered back projection reconstruction algorithms in 256-MDCT coronary angiography, AJR Am. J. Roentgenol. 199(3) (2012) 588-94. [22] A.M. den Harder, M.J. Willemink, R.P. Budde, A.M. Schilham, T. Leiner, P.A. de Jong, Hybrid and model-based iterative reconstruction techniques for pediatric CT, AJR. American journal of roentgenology 204(3) (2015) 645-53. [23] K.M. Brown, S. Zabic, T. Koehler, Acceleration of ML iterative algorithms for CT by the use of fast start images, Proc. SPIE 8313, Medical Imaging 2012: Physics of Medical Imaging, International Society for Optics and Photonics, 2012, pp. 831339-831339-7. [24] H. Yuki, D. Utsunomiya, Y. Funama, S. Tokuyasu, T. Namimoto, T. Hirai, R. Itatani, K. Katahira, S. Oshima, Y. Yamashita, Value of knowledge-based iterative model reconstruction in low-kV 256-slice coronary CT angiography, J. Cardiovasc. Comput. Tomogr. 8(2)

26

(2014) 115-23. [25] S.S. Son, K.S. Choo, U.B. Jeon, G.R. Jeon, K.J. Nam, T.U. Kim, J.A. Yeom, J.Y. Hwang, D.W. Jeong, S.J. Lim, Image quality of CT angiography with model-based iterative reconstruction in young children with congenital heart disease: comparison with filtered back projection and adaptive statistical iterative reconstruction, Int. J. Cardiovasc. Imaging 31 Suppl 1 (2015) 31-8. [26] Y. Yamada, M. Jinzaki, Y. Tanami, E. Shiomi, H. Sugiura, T. Abe, S. Kuribayashi, Model-based iterative reconstruction technique for ultralow-dose computed tomography of the lung: a pilot study, Invest. Radiol. 47(8) (2012) 482-9. [27] C.H. Park, J. Lee, C. Oh, K.H. Han, T.H. Kim, The feasibility of sub-millisievert coronary CT angiography with low tube voltage, prospective ECG gating, and a knowledge-based iterative model reconstruction algorithm, Int. J. Cardiovasc. Imaging 31 Suppl 2 (2015) 197-203. [28] P. Nie, H. Li, Y. Duan, X. Wang, X. Ji, Z. Cheng, A. Wang, J. Chen, Impact of sinogram affirmed iterative reconstruction (SAFIRE) algorithm on image quality with 70 kVp-tube-voltage dual-source CT angiography in children with congenital heart disease, PloS one 9(3) (2014) e91123. [29] M. Nakagawa, Y. Ozawa, K. Sakurai, M. Shimohira, K. Ohashi, M. Asano, S. Yamaguchi, Y. Shibamoto, Image quality at low tube voltage (70 kV) and sinogram-affirmed iterative reconstruction for computed tomography in infants with congenital heart disease, Pediatr. Radiol. 45(10) (2015) 1472-9.

27

[30] M. Asano, T. Saito, N. Nomura, A. Mishima, Double-outlet technique for tetralogy of Fallot-type disease with an anomalous coronary artery, Pediatr. Cardiol. 26(5) (2005) 710-2. [31] I.C. Tsai, T. Lee, M.C. Chen, Y.C. Fu, S.L. Jan, C.C. Wang, Y. Chang, Visualization of neonatal coronary arteries on multidetector row CT: ECG-gated versus non-ECG-gated technique, Pediatr. Radiol. 37(8) (2007) 818-25. [32] J. Leipsic, T.M. Labounty, B. Heilbron, J.K. Min, G.B. Mancini, F.Y. Lin, C. Taylor, A. Dunning, J.P. Earls, Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography, AJR Am. J. Roentgenol. 195(3) (2010) 649-54. [33] F.A. Mieville, F. Gudinchet, E. Rizzo, P. Ou, F. Brunelle, F.O. Bochud, F.R. Verdun, Paediatric cardiac CT examinations: impact of the iterative reconstruction method ASIR on image quality--preliminary findings, Pediatr. Radiol. 41(9) (2011) 1154-64.

28

Figure legends Fig. 1. A five-month-old boy with a ventricular septal defect and a dilated pulmonary artery. A 256-slice multi-detector computed tomography (MDCT) angiography scan was performed at 80 kV and 30 mAs/rotation (effective radiation dose=0.62 mSv). Axial images reconstructed using the filtered back projection (FBP) (A), iDose4 (B) and iterative model reconstruction (IMR) (C) algorithms are shown. The IMR algorithm yielded substantially reduced image noise and increased image quality compared with the FBP and iDose4 algorithms. Fig. 2. The same patient as shown in Fig. 1. Multi-planar reformatted (MPR) images of the left anterior descending (LAD) coronary artery reconstructed using the filtered back projection (FBP) (A), iDose4 (B) and iterative model reconstruction (IMR) (C) algorithms are shown. Images reconstructed using the IMR algorithm (C) showed lower image noise than did those reconstructed with the FBP (A) and iDose4 algorithms (B). IMR showed the left main (LM) coronary artery and the proximal segment of the LAD more clearly than did the FBP and iDose4 algorithms (arrow). However, IMR showed the middle segment of the LAD less clearly than did FBP and iDose4 (arrow head). Fig. 3. A one-month-old boy with transposition of the great arteries, a double-outlet right ventricle and patent ductus arteriosus (PDA). A 256-slice multi-detector computed tomography (MDCT) angiography scan was performed at 80 kV and 20 mAs/rotation (effective radiation dose=0.40 mSv). Sagittal multi-planar reformatted (MPR) images reconstructed using the filtered back projection (FBP) (A), iDose4 (B) and iterative model reconstruction (IMR) (C) algorithms are shown. FBP and iDose4 clearly showed the small PDA (arrow). However, the lesion was blurred on the MPR image processed using the IMR algorithm

29

Tables Table 1. Patient demographics, scanning protocols and radiation doses Table 2. Comparison of attenuation values, noise, CNR and SNR Table 3. Assessment of subjective image quality Table 4. Number of detected coronary artery segments (%) Table 5. Findings of 256-slice multi-detector computed tomography angiography using filtered back projection (FBP), iDose4 or iterative model reconstruction (IMR).

30

Patient demographics, scanning protocols and radiation doses Number, or mean ±SD (range) Demographics Patients

51

No. of male

29

Age (months)

4.36 ±4.09 (0.03-12)

Body weight (kg)

5.64 ±2.10 (2.20-11.20)

Heart rate during scan (beats/min)

137.86 ±21.34 (92-179)

Protocols Tube potential (kVp)

80

Tube current (mAs) 0-3 (Kg)

15

4 (Kg)

20

5 (Kg)

25

31

6 (Kg)

30

7 (Kg)

33

8 (Kg)

35

9 (Kg)

38

10Kg

40

11Kg

43

Radiation doses CTDIvol (mGy)

1.59 ±0.49(0.15-2.59)

DLP (mGy cm)

18.27 ±7.57(6.00-35.10)

Scan length (mm)

109.02 ±17.81(68.40-146.62)

ED (mSv)

0.61 ±0.32(0.16-1.37)

Note: Data are the mean ± SD. CTDIvol=Volume CT Dose Index, DLP=Dose Length Product, ED=Effective Dose

32

Comparison of attenuation values, noise, CNR and SNR Regions of interest

Reconstruction algorithm

P value

FBP

iDose4

IMR

Ascending aorta

600.59±149.66

597.69±150.14

605.88±150.86

0.96

Descending aorta

604.76±154.28

599.53±155.12

619.90±163.48

0.80

Main pulmonary artery

628.90±145.70

626.65±145.12

624.42±143.87

0.99

Right atrium

604.34±155.83

602.88±155.80

603.38±155.28

1.00

Right ventricle

600.65±140.56

618.10±236.53

597.18±132.97

0.81

Left atrium

596.02±163.36

594.80±163.57

598.62±164.66

0.99

Left ventricle

602.88±159.05

600.84±158.85

599.76±158.79

1.00

Ascending aorta

60.40±15.55

44.45±14.52

25.06±12.92

0.00*

Descending aorta

58.27±17.26

45.03±12.60

26.66±13.32

0.00*

Attenuation, HU

Noise

33

Main pulmonary artery

57.89±12.90

42.02±10.78

23.82±14.22

0.00*

Right atrium

92.28±44.48

76.13±46.45

56.57±42.70

0.00*

Right ventricle

62.21±12.94

46.07±12.32

30.99±33.05

0.00*

Left atrium

67.42±12.82

48.65±12.07

26.54±13.14

0.00*

Left ventricle

63.50±11.98

45.21±9.70

25.19±9.18

0.00*

Ascending aorta

10.38±3.11

14.54±5.65

27.92±10.26

0.00*

Descending aorta

11.57±6.04

13.94±3.94

27.82±12.58

0.00*

Main pulmonary artery

11.37±3.48

15.72±4.69

30.94±11.61

0.00*

Right atrium

7.21±2.53

9.37±4.18

15.36±13.95

0.00*

Right ventricle

10.00±2.97

14.17±6.25

24.85±10.26

0.00*

Left atrium

8.93±2.12

12.45±2.86

25.31±8.27

0.00*

Left ventricle

9.65±2.38

13.67±3.77

26.57±10.67

0.00*

11.85±4.16

17.50±6.38

44.03±14.72

0.00*

SNR

CNR Ascending aorta

34

Descending aorta

11.92±4.14

17.56±6.39

45.16±15.52

0.00*

Main pulmonary artery

12.49±4.30

18.51±6.53

45.51±13.41

0.00*

Right atrium

12.07±4.69

17.89±7.12

44.49±17.18

0.00*

Right ventricle

11.81±3.68

18.04±7.71

43.28±13.17

0.00*

Left atrium

11.99±5.29

17.79±8.07

44.17±17.33

0.00*

Left ventricle

12.01±4.69

17.82±7.20

43.89±15.24

0.00*

Note: Data are the mean ± SD. FBP = filtered back-projection, IMR = iterative model reconstruction, CNR= contrast-to-noise ratio, SNR=signal-to-noise ratio. *significant differences for comparison among the three methods.

35

Assessment of subjective image quality Objective image quality

Reconstruction algorithm

P value

FBP

iDose4

IMR

Image noise

2.49±0.50

3.49±0.50

4.41±0.50

0.000*

Margin sharpness

2.78±0.42

3.78±0.42

4.48±0.58

0.000*

Overall image quality

2.35±0.48

3.35±0.48

4.20±0.57

0.000*

Note: Data are the mean ± SD. FBP = filtered back-projection, IMR = iterative model reconstruction. *significant differences for comparison among the three methods.

36

Numbers of

detected Reconstruction algorithm

coronary

P value

FBP

iDose4

IMR

Total segments

229(64.15%)

267(74.79%)

257(71.99%)

0.006*

Left main coronary artery

48(94.12%)

49(96.08%)

49(96.08%)

0.862

Left anterior descending coronary artery

73(71.57%)

87(85.29%)

81(79.41%)

0.056

Left circumflex coronary artery

41(40.20%)

54(52.94%)

52(50.98%)

0.147

Right coronary artery

67(65.68%)

77(75.49%)

78(76.47%)

0.163

segments (%)

Note: Data are the mean ± SD. FBP = filtered back-projection, IMR = iterative model reconstruction. *significant differences for comparison among the 3 methods.

artery

37

Findings by filtered back projection (FBP), iDOSE4 and iterative model reconstruction (IMR) at 256-slice multi-detector CT angiography

deformities

iDdose4

FBP

Cardiovascular

IMR

Surgical/CCA

TP

TN

FP

FN

TP

TN

FP

FN

TP

TN

FP

FN

results

ASD

38

9

1

3

39

8

2

2

38

7

3

3

41

VSD

46

3

0

2

46

3

0

2

46

2

1

2

48

RVOTS

7

44

0

0

7

44

0

0

7

44

0

0

7

DORV

3

48

0

0

3

48

0

0

3

48

0

0

3

PA

3

48

0

0

3

48

0

0

3

48

0

0

3

PAS

14

37

0

0

14

37

0

0

14

37

0

0

14

DPA

19

32

0

0

19

32

0

0

19

32

0

0

19

APVR

9

42

0

0

9

42

0

0

9

42

0

0

9

PDA

11

39

0

1

11

39

0

1

10

39

0

2

12

PTA

1

50

0

0

1

50

0

0

1

50

0

0

1

Overriding aorta

12

39

0

0

12

39

0

0

12

39

0

0

12

38

COA

7

44

0

0

7

44

0

0

7

44

0

0

7

Right aortic arch

6

45

0

0

6

45

0

0

6

45

0

0

6

TGA

5

46

0

0

5

46

0

0

5

46

0

0

5

DSVC

4

47

0

0

4

47

0

0

4

47

0

0

4

CAA

2

49

0

0

2

49

0

0

2

49

0

0

2

Total

187

622

1

6

188

621

2

5

186

619

4

7

193

ASD=Atrial septal defect, VSD=Ventricular septal defect, RVOTS=Right ventricular outflow tract stenosis, DORV=Double outlet right ventricle, PA=Pulmonary artery atresia, PAS=Pulmonary artery stenosis, DPA=Dilated pulmonary artery, APVR=Anomalous pulmonary venous return, PDA=Patent ductus arteriosus, PTA=Persistent truncus arteriosus, COA=Coarctation of the aorta, TGA=Transposition of the great arteries, DSVC=Double superior vena cava, CAA=Coronary artery anomaly, TP= true positive detection, TN=true negative detection, FP=false positive detection, FN=false negative detection, CCA= Conventional cardiac angiography.