Performance of adaptive iterative dose reduction 3D integrated with automatic tube current modulation in radiation dose and image noise reduction compared with filtered-back projection for 80-kVp abdominal CT: Anthropomorphic phantom and patient study

European Journal of Radiology 85 (2016) 1666–1672

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Performance of adaptive iterative dose reduction 3D integrated with automatic tube current modulation in radiation dose and image noise reduction compared with filtered-back projection for 80-kVp abdominal CT: Anthropomorphic phantom and patient study Chien-Ming Chen (MD) a,b , Yang-Yu Lin (MD) a,b , Ming-Yi Hsu (MD) a,b , Chien-Fu Hung (MD) a,b , Ying-Lan Liao (PhD) c , Hui-Yu Tsai (PhD) c,d,∗ a

Department of Medical Imaging and Intervention, Chang Gung Memorial Hospital Linkou, 5 Fuxing Street, Kwei-Shan 333, Taoyuan, Taiwan College of Medicine, Chang Gung University, 259 Wen-Hwa 1st Road, Kwei-Shan 333, Taoyuan, Taiwan c Medical Physics Research Center, Institute for Radiological Research, Chang Gung University/Chang Gung Memorial Hospital, Linkou, 259 Wen-Hwa 1st Road, Kwei-Shan 333, Taoyuan, Taiwan d Department of Medical Imaging & Radiological Sciences, Chang Gung University, 259 Wen-Hwa 1st Road, Kwei-Shan 333, Taoyuan, Taiwan b

a r t i c l e

i n f o

Article history: Received 28 April 2016 Received in revised form 4 July 2016 Accepted 11 July 2016 Keywords: Multidetector computed tomography Radiation dosage Radiographic image enhancement Phantoms Imaging Observer variation

a b s t r a c t Objectives: Evaluate the performance of Adaptive Iterative Dose Reduction 3D (AIDR 3D) and compare with filtered-back projection (FBP) regarding radiation dosage and image quality for an 80-kVp abdominal CT. Materials and methods: An abdominal phantom underwent four CT acquisitions and reconstruction algorithms (FBP; AIDR 3D mild, standard and strong). Sixty-three patients underwent unenhanced liver CT with FBP and standard level AIDR 3D. Further post-acquisition reconstruction with strong level AIDR 3D was made. Patients were divided into two groups (< and 29 cm) based on the abdominal effective diameter (Deff ) at T12 level. Quantitative (attenuation, noise, and signal-to-noise ratio) and qualitative (image quality, noise, sharpness, and artifact) analysis by two readers were assessed and the interobserver agreement was calculated. Results: Strong level AIDR 3D reduced radiation dose by 72% in the phantom and 47.1% in the patient study compared with FBP. There was no difference in mean attenuations. Image noise was the lowest and signal-to-noise ratio the highest using strong level AIDR 3D in both patient groups. For Deff < 29 cm, image sharpness of FBP was significantly different from those of AIDR 3D (P < 0.05). For Deff 29 cm, image quality of AIDR 3D was significantly more favorable than FBP (P < 0.05). Interobserver agreement was substantial. Conclusions: Integrated AIDR 3D allows for an automatic reduction in radiation dose and maintenance of image quality compared with FBP. Using AIDR 3D reconstruction, patients with larger abdomen circumference could be imaged at 80 kVp. © 2016 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Abbreviations: AIDR 3D, adaptive iterative dose reduction 3D; ATCM, automatic tube current modulation; CTDIvol , CT volume dose index; Deff , effective diameter; DLP, dose length product; FBP, filtered-back projection; IR, iterative reconstruction; ROI, region of interest; SNR, signal-to-noise ratio; SSDE, size-specific dose estimate. ∗ Corresponding author at: Department of Medical Imaging & Radiological Sciences, Chang Gung University, 259 Wen-Hwa 1st Road, Kwei-Shan 333, Taoyuan, Taiwan. E-mail addresses: [email protected] (C.-M. Chen), [email protected] (Y.-Y. Lin), [email protected] (M.-Y. Hsu), [email protected] (C.-F. Hung), [email protected] (Y.-L. Liao), [email protected] (H.-Y. Tsai). http://dx.doi.org/10.1016/j.ejrad.2016.07.002 0720-048X/© 2016 Elsevier Ireland Ltd. All rights reserved.

Reducing the radiation dose from computed tomography (CT) exams is an area of active research because of concerns about the detrimental effects of radiation on patients [1]. Reducing the tube voltage and using automatic tube current modulation (ATCM) are established methods for minimizing CT radiation exposure [2]. Achievable dose reductions are in the order of 20%–50% for using ATCM [3,4] and 24%–48% for low tube voltage acquisitions [5]. However, until the introduction of iterative reconstruction (IR), low tube voltage resulted in increased image noise, thus limiting

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

its use. As opposed to filtered-back projection (FBP), IR techniques incorporate a physical model of the CT system that more accurately reproduce the data acquisition process [6]. IR can be subclassified into two major categories: (1) hybrid reconstruction that involves blending of FBP with IR images and (2) pure or model-based reconstruction in the space domain [6]. A new generation of IR that works on both projection and image space data has shown greater ability in reducing noise [7,8]. One such hybrid reconstruction method is adaptive iterative dose reduction 3D (AIDR 3D, Toshiba Medical Systems, Otawara, Japan), which has been integrated into the imaging chain through ATCM (SURE Exposure 3D) and affects both image noise and radiation exposure through tube current reduction [9,10]. AIDR 3D is available at three strength levels: strong, standard, and mild [11]. The performance of the different strengths has not been previously studied. Other hybrid and pure IR techniques introduced by CT vendors include adaptive statistical iterative reconstruction(ASIR)/model-based iterative reconstruction(MBIR) (GE Healthcare, Waukesha, USA), sinogram affirmative iterative reconstruction (SAFIRE)/adaptive model iterative reconstruction(ADMIRE) (Siemens Healthcare, Erlangen, Germany) and iDose4 /iterative model reconstruction(IMR) (Philips Healthcare, Best, the Netherlands). With the advent of IR, imaging of patients at reduced tube voltages has become practical even in larger size patients due to its significant image noise reducing capability [12]. Several studies on IR have reported potential radiation dose reductions when extrapolating reductions in image noise with IR during post-processing [13,14]. As comparing with reference standard FBP algorithm, other studies have achieved reductions in image noise and radiation dose through estimated or calculated manual adjustment of the image quality indicator (i.e., the noise index) for IR [10,15–17]. However, a recent liver phantom study showed that despite the increased contrast-to-noise ratio of AIDR 3D images, there was a lower sensitivity for low-contrast lesion detection when the radiation dose was reduced to 20% of the reference [18]. Similarly, another study showed that for middle-contrast objects, the modulation transfer function of AIDR 3D decreased with decreasing radiation dose and increasing strength of AIDR 3D [19]. For overcoming this, we hypothesized that using the same noise index setting as FBP for AIDR 3D would result in immediate radiation dose reduction and image quality comparable with those of reduced voltage abdominal CT. Through validation with anthropomorphic phantom and clinical patients, this study evaluated the performance of AIDR 3D when integrated into ATCM compared with FBP reconstruction by evaluating the radiation dose, image quality, and image noise for 80-kVp abdominal CT. 2. Materials and methods 2.1. Study design This prospective study was approved by the Institutional Review Board. Prior to the patient study, we conducted a phantom study to validate the effects of AIDR 3D integration into ATCM. In the patient study, we enrolled patients for follow-up CT after colon cancer treatment. Intraindividual comparisons were made to evaluate diagnostic image quality and radiation dose from FBP versus AIDR 3D. 2.2. Phantom study We used a commercial abdominal phantom (Model 057 Triple Modality 3D Abdominal Phantom, CIRS, VA, USA) with simplified anthropomorphic geometry (width: length: height,

1667

26 cm × 12.5 cm × 19 cm) to simulate the abdominal region from the T9/T10 to L2/L3 vertebra. The internal structure of the phantom includes the liver, kidneys, lung, abdominal aorta, spine, muscle, and outside fat layer. 2.3. Patient study The Radiology Information System was used to identify patients scheduled for a CT examination. Inclusion criteria were the following: age  18 years, weight <90 kg, the ability to provide written informed consent, and the ability to hold one’s breath while remaining still for at least 10 s. Between February 1, 2013 and May 31, 2013, 63 consecutive patients were identified; all of them participated in the study. The age, sex, weight, and height of each patient was recorded. The body-mass index (BMI) was calculated as the weight divided by the square of the height. 2.4. CT acquisition CT acquisition was performed on a 320-row multidetector CT (Aquilion ONE, Toshiba Medical Systems; software Version 4.74ER004). To assess the performance of the ATCM and reconstruction algorithms, the entire length of the phantom was acquired four times with four reconstruction algorithms (FBP; AIDR 3D mild, standard, and strong). For the patient study, to minimize additional radiation exposure, we focused on the unenhanced liver CT. Two unenhanced CT acquisitions of equal length covering the whole liver were performed while the patient held one breath. The first acquisition involved using FBP and the second an AIDR3D standard algorithm (based on the findings of the phantom study). The other imaging parameters are as follows: the acquisition mode, helical; detector collimator dimensions, 80 × 0.5 mm; tube potential, 80 kVp; gantry rotation time, 0.75 s; table pitch, 0.638; x,y,z-axis tube current modulation (SURE Exposure 3D); standard deviation of noise, 9; tube current, 10–580 mA; reconstruction kernel, FC18; and slice thickness and interval, 5 mm. 2.5. Image reconstruction Four image sets were reconstructed (FBP and varying strengths of AIDR 3D) for the phantom study. For the patient study, a second set of images was reconstructed using the AIDR 3D strong algorithm from raw data obtained in the AIDR 3D acquisition. The subjective diagnostic acceptability of the two types of IR was evaluated and the objective image noise differences were determined. In total, three sets of images (FBP, AIDR 3D standard, and AIDR 3D strong) were analyzed. 2.6. Quantitative image analysis Circular regions of interest (ROI) were made for each image set at the right and left lobe of the liver, the aorta at the level of the portal vein, and the psoas muscle in both the phantom and the patient. For the liver, ROIs were made as large as possible over a homogenous region and care was taken to avoid vessels and calcifications. In the psoas muscle, the surrounding bone and fat were carefully avoided. The mean attenuation value (in Hounsfield Unit, HU) and standard deviation (representing the noise) of each ROI were recorded. To ensure consistency in the phantom, the ROIs were copied and pasted to subsequent acquisitions. To minimize bias from a single measurement, we calculated the average of all ROI measurements at three consecutive slices. For each specific ROI, the signal-to-noise ratio (SNR) was calculated by dividing the mean attenuation value by the standard deviation. Patients were subdivided into two subgroups (<29 cm and 29 cm) based on their calculated effective diameter (Deff ) to facilitate image analysis. A

1668

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

Deff of 29 cm is approximately equal to a waist circumference of 90 cm—the cut-off value for obesity in the study population [20]. The Deff calculations are explained subsequently in the section of radiation dose estimation. 2.7. Qualitative image analysis We performed a qualitative assessment on patients that was based on established recommendations. Two radiologists (MYH and YYL, with 5 and 2 years of abdominal imaging experience, respectively) reviewed all the images separately and independently. The CT images were randomized, and the readers were blinded to the image acquisition parameters. Images were presented with a preset window level of 70 and window width of 400, but readers were free to make adjustments. Before formal evaluation, the criteria for image grading were established by the readers and a standard reference image was selected for each grade. These images were not included in the formal reading. Readers graded the images for overall image quality (1–inadequate for diagnosis; 2–acceptable; 3–adequate; 4–optimal), noise (1–very noisy, unacceptable; 2–average noise; 3–less than average noise; 4–minimum or no noise), sharpness (1–very blurry, unacceptable; 2–acceptable; 3–adequate; 4–very sharp), and artifacts (1–unacceptable; 2–mild artifacts; 3–negligible artifacts; 4–no artifact). 2.8. Radiation dose estimation The CT volume dose index (CTDIvol ) and dose length product (DLP) shown on the console were verified using dose measurements and have a difference of under 10%. The CTDIvol and DLP of each CT acquisition on the console were recorded. Although the CTDIvol can be used to track changes in the patient radiation dose as methods change, it does not reflect changes in the patient dose because of a change in patient size. Because the patient dose is increased as the patient size decreases for a constant radiation output, underestimation of the patient dose occurs when the CTDIvol is used as the patient dose. To avoid this pitfall, we used an improved estimate of the patient dose, the size-specific dose estimate (SSDE), which is based on body size [21]. For the patient study, the anterioposterior (AP) and lateral (LAT) dimensions of the patient at the level of T12 vertebra body were measured from the topogram, and √ the effective diameter was calculated (Deff = (AP * LAT). The SSDE was then calculated from CTDIvol and Deff , by using the appropriate conversion factor. 2.9. Statistical analysis Statistical analysis was performed using statistical software (Statistica, Version 7.1; www.statsoft.com). We presented continuous variables (i.e., age, weight, height, BMI, attenuation, image noise, SNR, CTDIvol , DLP, and SSDE) as means ± standard deviations. The two patient groups were compared using the two-tailed Student’s t test for normally distributed data, or the Wilcoxon signed-rank test for nonnormally distributed data. The three image data sets were compared using a one-way ANOVA (analysis of variance), and Bonferroni and Tamhane’s T2 post hoc tests were used when variances had different Levene’s test values. Ordinal variables (i.e., image quality, subjective noise, sharpness, and artifacts) were presented as modes with frequency and were compared using the nonparametric Kruskall-Wallis ANOVA; P < 0.05 indicates a significant difference. Interobserver agreement was assessed using Cohen’s weighted kappa test. The degree of interobserver agreement for each qualitative assessment (image quality, noise, sharpness, and artifacts) was determined by calculating the ␬ value. The scale of ␬ coefficients for interobserver agreement was as follows: a ␬ value of less than 0.20

Table 1 Phantom results. AIDR 3D strong AIDR 3D standard AIDR 3D mild FBP Radiation dose 3.0 CTDIvol (mGy) DLP (mGy.cm) 52.6 Dose reduction (%) 72

3.0 52.6 72

6.0 105.6 38

11.9 209

Mean attenuation (HU) 88.6 Right lobe 92.9 Left lobe 84.9 Aorta Muscle 40.2

88.7 92.9 84.8 40.4

88.4 92.4 85.5 39.5

89.7 93.3 85.1 39.0

Image noise (HU) Mean total Right lobe Left lobe Aorta Muscle

12.0 12.3 11.1 12.4 12.0

13.7 14.3 12.8 14.1 13.5

12.3 13.4 10.3 12.1 13.4

12.9 14.1 10.5 14.4 12.4

SNR Mean total Right lobe Left lobe Aorta Muscle

6.5 7.2 8.4 6.8 3.4

5.6 6.2 7.3 6.0 3.0

6.4 6.6 9.0 7.1 2.9

6.1 6.4 8.9 5.9 3.1

FBP filtered-back projection, AIDR 3D adaptive iterative dose reduction 3D, CTDIvol CT volume dose index, DLP dose length product, SNR signal-to-noise ratio.

indicated poor agreement; a ␬ value of 0.21–0.40, fair agreement; a ␬ value of 0.41–0.60, moderate agreement; a ␬ value of 0.61–0.80, substantial agreement; and a ␬ value of 0.81–1.00, almost perfect agreement. 3. Results 3.1. Phantom study Table 1 presents a summary of the radiation dose and quantitative analysis of the phantom study. When compared with FBP, the AIDR 3D mild algorithm showed a 38% dose reduction, whereas both the AIDR 3D standard algorithm and strong algorithm showed a 72% dose reduction. The AIDR 3D strong algorithm showed a 12.5% reduction in mean image noise compared with the AIDR 3D standard algorithm. 3.2. Patient study Table 2 presents a summary of the patient demographics. There were 43 males and 20 females with a mean age of 60.6 ± 12.1 years (range 19–89 years). Except for the BMI (P < 0.01), there was no statistical difference in the age, weight, and height between the Deff < 29-cm and 29-cm subgroups. 3.3. Quantitative image analysis Table 3 presents a summary of the quantitative analysis of the patient study. The mean attenuation of the ROIs at the right and left lobe of the liver, aorta, and psoas muscle were not statistically significant for FBP or the AIDR 3D standard and strong algorithm in the study cohort (P = 0.19–0.99). The mean image noise was lowest with the AIDR 3D strong algorithm in the whole study cohort and in the two Deff subgroups (all P < 0.01). The mean noise reduction across all ROIs with AIDR 3D standard and strong algorithms compared with FBP were 23.2% vs 34.6%, 16.2% vs 28.5%, and 35.2% vs 45.1%, respectively. Similarly, the SNR was consistently more favorable with the AIDR 3D standard and strong algorithm compared with that of FBP (all P < 0.01).

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

1669

Table 2 Patient demographics.

Sex ratio (M:F) Age (years) Weight (kg) Height (cm) BMI (kg/m2 )

Study cohort

Effective diameter <29 cm

Effective diameter 29 cm

P value

43:20 60.6 ± 12.1 (19–89) 64.1 ± 10.0 (42–90) 161.3 ± 7.3 (144–176) 24.6 ± 3.3 (18.2–37.8)

22:18 59.0 ± 12.9 (19–89) 59.0 ± 7.6 (42–73) 160.0 ± 7.9 (144–176) 23.0 ± 2.3 (18.2–28.9)

21:2 63.5 ± 10.1 (47–83) 73.0 ± 7.0 (62–90) 163.7 ± 5.7 (150–174) 27.4 ± 3.1 (23.3–37.8)

0.22 0.72 0.10 <0.01

Data presented as mean ± standard deviation with range in parentheses. BMI: body-mass index.

3.4. Qualitative image analysis Tables 4 and 5 summarize the qualitative analysis for the patient subgroups Deff < 29 cm and Deff  29 cm, respectively. FBP and the AIDR 3D standard and strong algorithm in patients with Deff < 29 cm had significant differences in image sharpness and artifacts (P < 0.01), but not in image quality and noise. By contrast, in patients with Deff  29 cm, there was a statistically significant difference in image quality, noise, and artifacts (all P < 0.05). For Deff  29 cm, both readers rated the AIDR 3D strong algorithm to have a lower image noise compared with the AIDR 3D standard algorithm (P < 0.05), keeping with the results of quantitative image analysis. Except for image sharpness in patients with Deff < 29 cm, over which the readers only expressed moderate agreement (␬ = 0.50), there was substantial agreement in all other parameters. 3.5. Radiation dose estimation Table 6 shows the radiation dose for the patient study. The average dose reduction with AIDR 3D algorithm in the Deff < 29-cm and 29-cm subgroups was 55.3% and 32.8%, respectively. For patients with Deff  29 cm, the upper limit of CTDIvol (20.2 mGy) of this protocol was reached during FBP acquisitions. The mean SSDE in the study cohort was 25.2 ± 1.4 mGy and 13.3 ± 2.7 mGy for FBP and AIDR 3D, respectively (P < 0.01). 4. Discussion Integration of AIDR 3D into ATCM reduces the radiation dose during CT acquisitions through tube current reduction and reduces

image noise during image reconstruction, resulting in subjective and objective improvements in image quality compared with those of FBP. This improvement is most significant in patients with Deff  29 cm (Fig. 1). Our results agree with the phantom results of Kim et al. [22], who showed higher image quality in AIDR 3D images compared with those of FBP in 30 cm and 40 cm versus a 24-cm phantom. They surmised that the effect of the AIDR 3D algorithm increases as a patient’s body size increases. This effect is caused by the application of filter strength adaptation to the relative noise level in the projection space and also by the adaptive, weighted, anisotropic diffusion de-noising in image space [11,22]. The resulting image has a substantially reduced amount and slightly different distribution of noise (i.e., the noise power spectrum) but no deterioration in spatial resolution compared with FBP [23]. The phantom results showed that the radiation doses from using AIDR 3D standard and strong algorithm acquisitions are the same. Quantitatively, AIDR 3D strong algorithm reduced the image noise by an additional 10% compared with that of AIDR 3D standard algorithm. This reduction was qualitatively significant in patients with Deff  29 cm. No loss of image sharpness was caused by greater noise reduction. Lesion conspicuity and overall image quality improved with AIDR 3D strong algorithm compared with AIDR 3D standard algorithm in the 30 cm phantom study by Kim et al. [22]. Based on these results, AIDR 3D strong algorithm should be used in patients with Deff  29 cm to mitigate higher image noise. In patients with a Deff < 29 cm, there was no significant difference in the subjective image quality or image noise, but there was a loss of subjective sharpness in AIDR 3D images compared with FBP images (Fig. 2). Both readers rated the images with AIDR 3D standard and strong algorithms as adequate sharpness. The image noise for FBP and AIDR 3D images were rated equally to have

Table 3 Patient quantitative analysis. Study cohort FBP

Effective diameter 29 cm

Effective diameter <29 cm AIDR 3D standard AIDR 3D strong FBP

AIDR 3D standard AIDR 3D strong FBP

AIDR 3D standard AIDR 3D strong

Mean attenuation (HU) Right lobe livera 54.4 ± 13.3 56.2 ± 12.3 Left lobe livera 50.2 ± 4.7 Aortaa 52.7 ± 5.9 Psoas musclea

53.7 ± 13.2 55.6 ± 12.0 49.5 ± 5.6 50.8 ± 5.9

54.2 ± 13.0 56.0 ± 11.9 49.8 ± 5.5 51.6 ± 5.9

56.5 ± 11.7 59.5 ± 10.6 49.4 ± 4.4 53.4 ± 4.7

56.1 ± 11.6 58.8 ± 10.4 48.9 ± 5.7 51.1 ± 5.1

56.5 ± 11.5 59.1 ± 10.2 49.2 ± 5.5 51.7 ± 5.1

50.6 ± 15.1 50.5 ± 13.2 51.7 ± 4.8 51.4 ± 7.5

49.6 ± 15.0 50.2 ± 12.9 50.7 ± 5.5 50.2 ± 7.1

50.1 ± 14.8 50.5 ± 12.8 50.9 ± 5.3 51.4 ± 7.3

Mean image noise (HU) Right lobe liverb 17.6 ± 5.5 15.4 ± 4.4 Left lobe liverb 19.2 ± 5.7 Aortab 19.5 ± 6.1 Psoas muscleb

13.1 ± 2.2 12.4 ± 1.8 13.5 ± 2.1 14.6 ± 2.7

11.2 ± 1.9 10.6 ± 1.6 11.5 ± 1.8 12.2 ± 2.2

14.7 ± 3.2 12.9 ± 2.5 15.9 ± 2.9 16.4 ± 3.8

12.2 ± 1.3 11.6 ± 0.9 12.6 ± 1.1 13.5 ± 1.8

10.4 ± 1.2 9.9 ± 1.1 10.9 ± 1.0 11.3 ± 1.5

22.8 ± 4.9 19.8 ± 3.5 24.9 ± 4.8 24.9 ± 5.6

14.8 ± 2.4 13.9 ± 1.9 15.0 ± 2.4 14.6 ± 2.7

12.7 ± 2.2 11.8 ± 1.8 12.7 ± 2.2 13.8 ± 2.4

Mean SNR Right lobe liverb Left lobe liverb Aortab Psoas muscleb

4.2 ± 1.2 4.6 ± 1.2 3.8 ± 0.6 3.6 ± 0.7

5.0 ± 1.4 5.4 ± 1.4 4.4 ± 0.8 4.4 ± 0.09

4.1 ± 1.2 4.8 ± 1.3 3.2 ± 0.7 3.4 ± 0.9

4.6 ± 1.0 5.1 ± 1.0 3.9 ± 0.6 3.9 ± 0.7

5.5 ± 1.2 6.0 ± 1.1 4.6 ± 0.7 4.7 ± 0.8

2.3 ± 1.0 2.6 ± 0.9 2.1 ± 0.4 2.2 ± 0.5

3.4 ± 1.2 3.6 ± 0.9 3.5 ± 0.6 3.1 ± 0.6

4.0 ± 1.3 4.3 ± 1.1 2.8 ± 0.6 3.8 ± 0.8

3.4 ± 1.4 4.0 ± 1.6 2.8 ± 0.8 3.0 ± 1.0

Data presented as mean ± standard deviation. FBP filtered-back projection, AIDR 3D adaptive iterative dose reduction 3D, SNR signal-to-noise ratio. a Denotes no statistical significant difference. b Denotes statistical significant difference.

1670

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

Table 4 Patient qualitative analysis for effective diameter <29 cm. Reader 1

Image quality Noise Sharpness Artifact

␬

Reader 2

FBP

AIDR 3D standard

AIDR 3D strong

P value

FBP

AIDR 3D standard

AIDR 3D strong

P value

value

3 (0/6/22/12) 4 (0/6/12/22) 4 (0/0/18/22) 3 (0/4/30/6)

3 (0/10/20/10) 4 (0/10/12/18) 3 (0/2/34/4) 4 (0/4/10/26)

3 (0/8/22/10) 4 (0/0/18/22) 3 (0/4/24/12) 4 (0/0/14/26)

0.62 0.20 <0.05 <0.05

3 (0/6/24/10) 4 (0/6/12/22) 4 (0/2/16/22) 3 (0/4/30/6)

3 (0/2/26/12) 4 (0/10/10/20) 3 (0/0/36/4) 4 (0/4/10/26)

3 (0/8/26/6) 4 (0/0/18/22) 3 (0/8/28/4) 4 (0/0/14/26)

0.87 0.39 <0.05 <0.05

0.64 0.70 0.50 0.68

Data are modes with frequency of each score in parentheses. FBP filtered-back projection, AIDR 3D adaptive iterative dose reduction 3D.

Table 5 Patient qualitative analysis for effective diameter 29 cm. Reader 1

Image quality Noise Sharpness Artifact

␬

Reader 2

FBP

AIDR 3D standard

AIDR 3D strong

P value

FBP

AIDR 3D standard

AIDR 3D strong

P value

value

2 (6/13/4/0) 2 (4/10/5/4) 4 (0/2/9/11) 2 (6/9/7/1)

3 (1/6/12/4) 3 (1/6/9/7)a 3 (0/2/14/7) 3 (0/6/11/6)

3/4 (0/5/9/9) 3 (0/5/13/5)a 3 (0/5/13/5) 4 (0/4/9/10)

<0.05 0.04 0.19 <0.05

2 (3/15/5/0) 2 (3/11/9/0) 4 (0/0/9/14) 2 (7/11/5/0)

3 (1/8/12/2) 3 (0/7/10/6)a 3 (0/2/13/8) 3 (0/5/13/5)

4 (0/4/9/10) 3 (0/2/16/5)a 3 (0/5/12/6) 3 (0/5/10/8)

<0.05 <0.05 0.15 <0.05

0.63 0.64 0.61 0.78

Data are modes with frequency of each score in parenthesis. FBP filtered-back projection, AIDR 3D adaptive iterative dose reduction 3D. a Significant difference was demonstrated between AIDR 3D standard and AIDR 3D strong reconstructions by both readers.

Table 6 Patient radiation dose comparison. Study cohort

CTDIvol (mGy) DLP (mGy cm) SSDE (mGy) Effective diameter (cm) Dose reduction (%)

Effective diameter 29 cm

Effective diameter <29 cm

FBP

AIDR 3D

P value

FBP

AIDR 3D

P value

FBP

AIDR 3D

P value

18.8 ± 1.5 463.9 ± 52.7 25.2 ± 1.4

10.1 ± 3.0 250.5 ± 81.7 13.3 ± 2.7 28 ± 3 47.1 ± 12.7

<0.01 <0.01 <0.01

18.0 ± 1.4 441.0 ± 41.7 25.9 ± 1.2

8.1 ± 1.6 198.8 ± 41.3 11.6 ± 1.6 26 ± 2 55.3 ± 6.0

<0.01 <0.01 <0.01

20.0 ± 0.2 503.7 ± 46.0 24.2 ± 1.0

13.5 ± 1.6 340.4 ± 49.6 16.2 ± 1.3 31 ± 1 32.8 ± 7.3

<0.01 <0.01 <0.01

Data presented as mean ± standard deviation. FBP filtered-back projection, AIDR 3D adaptive iterative dose reduction 3D, CTDIvol CT volume dose index, DLP dose length product, SSDE size-specific dose estimate.

Fig. 1. 47 year-old male after colon cancer treatment follow up (weight 80 kg, BMI 27.7 kg/m2 , Deff of 30 cm). Unenhanced axial CT images at the level of the portal trunk with (a) FBP, (b) AIDR 3D standard algorithm, and (c) AIDR 3D strong algorithm. The SSDE was 24.6 mGy (maximum radiation dose output of 20 mGy was reached) for FBP and 15.9 mGy for AIDR 3D, representing a 35.4% reduction in radiation dose. Note that FBP reconstruction results in higher image noise than AIDR 3D. All three images were subjectively found to be similarly sharp.

Fig. 2. 56-year-old male after colon cancer treatment follow up (weight 53 kg, BMI 19.9 kg/m2 , Deff of 24 cm). Unenhanced axial CT images at the level of the portal trunk with (a) FBP, (b) AIDR 3D standard algorithm, and (c) AIDR 3D strong algorithm. The SSDE was 24.2 mGy for FBP and 9.6 mGy for AIDR 3D, representing a 60.1% reduction in radiation dose. Although the overall image quality of the AIDR 3D images was the same as FBP, there was a slight loss in image sharpness, particularly with AIDR 3D strong.

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

minimum image noise. The smoothing effect of noise reduction and the altered noise texture in AIDR 3D images may influence the perception of edge sharpness. A study comparing routine-dose FBP with low-dose AIDR 3D CT urography found that the images obtained from low-dose AIRD 3D has statistically worse image sharpness at the renal pelvis and urinary bladder [24]. However, the diagnostic acceptability was not affected. By contrast, in a study comparing routine-dose FBP and low-dose AIDR 3D abdominal CT, no significant difference was found in image sharpness [11]. In an intraindividual analysis of objective contour sharpness in coronary CT angiography, no significant difference was found between FBP and the three strengths of AIDR 3D [25]. In an 80-kVp abdominal CT with IR reconstruction, the attenuation difference between hypovascular liver tumor and hepatic parenchyma plays a critical role in maintaining diagnostic accuracy [26,27]. In such a case, image sharpness would not affect lesion detection. Reducing the tube voltage to 80 kVp has been previously studied for chest [28], liver and aorta [5] imaging. One of the major limitations in reducing the tube voltage is the resulting image noise which limits its use in larger patients [29]. Several authors have suggested upper limits of usage based on patient weight and size [30,31]. Although weight-based protocols are useful guidance for CT scanning protocol selection, the image quality may not correlate well with the region been imaged [29]. One recent study based on dual-energy CT showed that the longest linear dimension was more likely to result in an unacceptable image compared with other clinical size parameters (weight, lean body weight, and BMI) [29]. They suggested image quality becomes unacceptable in abdominopelvic CT at a lateral dimension of 36 cm. In our study, despite all patients weighing less than 90 kg, we showed that a Deff  29 cm (lateral dimension of near 34 cm) was the upper limit for acceptable image quality for FBP at 80 kVp. For low tube voltage examinations, AIDR 3D allows for higher tolerance of patient size and weight. In most patients with Deff  29 cm, the maximal radiation dose output was reached with FBP, whereas AIDR 3D acquisitions reached 70% of maximal output. This means that much larger and heavier patients could be imaged with low tube voltages. Future studies should determine the upper limits of patient parameters suitable for low tube voltage AIDR 3D protocols. Another most substantial benefit of low tube voltage is the increase in the image contrast for structures with high effective atomic number, such as calcium and iodine [32]. This study has several limitations. First, we focused on image quality and radiation dose, not diagnostic accuracy. In all followup patients, no liver lesions were detected upon a subsequent enhanced CT. Second, we did not evaluate the effect of the higher attenuation of calcified structures and iodinated contrast medium at 80 kVp. The photoelectric effect is accentuated at low voltages and would affect the Hounsfield unit measurements in post contrast studies and the standard window and level settings for image display [32]. However, our study findings on non-contrast CT images may serve as a basis for future work on contrast--enhanced CT applications.Third, we did not evaluate the effects from of the altered noise texture (i.e., the noise power spectrum) and edge smoothing of IR images on reader diagnostic confidence. Reader studies have shown strong-level ASIR have produce “plastic” or “wax-like” appearance of images, possibly due to the lack of image noise7 . In our present study both readers did not report such an appearance with AIDR 3D strong algorithm. Fourth, the results only apply to the specified protocol, CT scanner, and patient population. In conclusion, integration of AIDR 3D into ATCM allows for an automatic reduction in radiation dose while maintaining image quality at 80 kVp compared with FBP. Additionally, AIDR 3D allows much larger and heavier patients to be imaged at low tube voltages.

1671

Conflict of interest The authors have no conflicts of interest and no disclosures. Role of the funding source • Financial activities related to the present article: funding supported from Chang Gung Memorial Hospital. • Financial activities not related to the present article: none to disclose. • Other relationships: none to disclose. Acknowledgments This study was supported by the grant from Chang Gung Memorial Hospital (CMRPG1D0051, CMRPD1B0331, CIRPD1C0053). Hui-Yu Tsai is a research scientist supported by Chang Gung Memorial Hospital (BMRPA61). The authors thank Dose Assessment Core Laboratory of Institute for Radiological Research for help regarding dose assessment. The authors also acknowledge the assistance of the dedicated radiology staff at Chang Gung Memorial Hospital at Linkou. References [1] R. Smith-Bindman, J. Lipson, R. Marcus, et al., Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer, Arch. Intern. Med. 22 (2009) 2078–2086. [2] J. Valentin Managing patient dose in multi-detector computed tomography(MDCT). ICRP Publication 102. Ann ICRP. 1 (2007) 1–79, iii. [3] A.J. Van der Molen, R.M.S. Joemai, J. Geleijns, Performance of longitudinal and volumetric tube current modulation in a 64-slice CT with different choices of acquisition and reconstruction parameters, Phys. Med. 4 (2012) 319–326. [4] J.B. Solomon, X. Li, E. Samei, Relating Noise to image quality indicators in CT examinations with tube current modulation, Am. J. Roentgenol. 3 (2013) 592–600. [5] C.M. Chen, S.Y. Chu, M.Y. Hsu, Y.L. Liao, H.Y. Tsai, Low-tube-voltage (80 kVp) CT aortography using 320-row volume CT with adaptive iterative reconstruction: lower contrast medium and radiation dose, Eur. Radiol. 2 (2014) 460–468. [6] R.D.A. Khawaja, S. Singh, A. Otrakji, et al., Dose reduction in pediatric abdominal CT: use of iterative reconstruction techniques across different CT platforms, Pediatr. Radiol. 7 (2015) 1046–1055. [7] M.J. Willemink, P.A. De Jong, T. Leiner, et al., Iterative reconstruction techniques for computed tomography Part 1: technical principles, Eur. Radiol. 6 (2013) 1623–1631. [8] M.J. Willemink, T. Leiner, P.A. De Jong, et al., Iterative reconstruction techniques for computed tomography part 2: initial results in dose reduction and image quality, Eur. Radiol. 6 (2013) 1632–1642. [9] Y. Ohno, D. Takenaka, T. Kanda, et al., Adaptive iterative dose reduction using 3D processing for reduced- and low-dose pulmonary CT: comparison with standard-dose CT for image noise reduction and radiological findings, Am. J. Roentgenol. 4 (2012) W477–W485. [10] A. Gervaise, B. Osemont, M. Louis, S. Lecocq, P. Teixeira, A. Blum, Standard dose versus low-dose abdominal and pelvic CT: comparison between filtered back projection versus adaptive iterative dose reduction 3D, Diagn. Interv. Imaging 1 (2014) 47–53. [11] M. Matsuki, T. Murakami, H. Juri, S. Yoshikawa, Y. Narumi, Impact of adaptive iterative dose reduction (AIDR) 3D on low-dose abdominal CT: comparison with routine-dose CT using filtered back projection, Acta Radiol. 8 (2013) 869–875. [12] E.C. Ehman, L. Yu, A. Manduca, et al., Methods for clinical evaluation of noise reduction techniques in abdominopelvic CT, Radiographics 4 (2014) 849–862. [13] A. Winklehner, C. Karlo, G. Puippe, et al., Raw data-based iterative reconstruction in body CTA: evaluation of radiation dose saving potential, Eur. Radiol. 12 (2011) 2521–2526. [14] P. Prakash, M.K. Kalra, A.K. Kambadakone, et al., Reducing abdominal ct radiation dose with adaptive statistical iterative reconstruction technique, Invest. Radiol. 4 (2010) 202–210. [15] Y. Sagara, A.K. Hara, W. Pavlicek, A.C. Silva, R.G. Paden, Q. Wu, C.T. Abdominal, Comparison of low-dose CT with adaptive statistical iterative reconstruction and routine-dose CT with filtered back projection in 53 patients, Am. J. Roentgenol. 3 (2010) 713–719. [16] L.M. Mitsumori, W.P. Shuman, J.M. Busey, O. Kolokythas, K.M. Koprowicz, Adaptive statistical iterative reconstruction versus filtered back projection in the same patient: 64 channel liver CT image quality and patient radiation dose, Eur. Radiol. 1 (2012) 138–143.

1672

C.-M. Chen et al. / European Journal of Radiology 85 (2016) 1666–1672

[17] Y. Yamada, M. Jinzaki, T. Hosokawa, et al., Dose reduction in chest CT: comparison of the adaptive iterative dose reduction 3D adaptive iterative dose reduction, and filtered back projection reconstruction techniques, Eur. J. Radiol. 12 (2012) 4185–4195. [18] S.T. Schindera, D. Odedra, S.A. Raza, et al., Iterative reconstruction algorithm for CT: can radiation dose be decreased while low-contrast detectability is preserved, Radiology 2 (2013) 511–518. [19] A. Urikura, K. Ichikawa, T. Hara, E. Nishimaru, Y. Nakaya, Spatial resolution measurement for iterative reconstruction by use of image-averaging techniques in computed tomography, Radiol. Phys. Technol. 2 (2014) 358–366. [20] K.C. Huang, M.S. Lee, S.D. Lee, et al., Obesity in the elderly and its relationship with cardiovascular risk factors in Taiwan, Obes. Res. 1 (2005) 170–178. [21] J.M. Boone, K.J. Strauss, D.D. Cody, C.H. McCollough, M.F. McNitt-Gray, T.L. Toth, Size Specific Dose Estimates (SSDE) in Pediatric and Adult Body CT Examinations, American Association of Physicists in Medcine, 2011, pp. 1–30. [22] M. Kim, J.M. Lee, J.H. Yoon, et al., Adaptive iterative dose reduction algorithm in CT: effect on image quality compared with filtered back projection in body phantoms of different sizes, Korean J. Radiol. 2 (2014) 195–204. [23] R.M.S. Joemai, W.J.H. Veldkamp, L.J.M. Kroft, I. Hernandez-Giron, J. Geleijns, Adaptive iterative dose reduction 3d versus filtered back projection in CT: evaluation of image quality, Am. J. Roentgenol. 6 (2013) 1291–1297. [24] H. Juri, M. Matsuki, Y. Itou, et al., Initial experience with adaptive iterative dose reduction 3D to reduce radiation dose in computed tomographic urography, J. Comput. Assisted Tomogr. 1 (2013) 52–57. [25] S. Feger, M. Rief, E. Zimmermann, et al., The impact of different levels of adaptive iterative dose reduction 3D on image quality of 320-row coronary CT angiography: a clinical trial, PLoS One 5 (2015).

[26] E. Robinson, J. Babb, H. Chandarana, M. MacAri, Dual source dual energy MDCT: comparison of 80 kVp and weighted average 120 kVp data for conspicuity of hypo-vascular liver metastases, Invest. Radiol. 7 (2010) 413–418. [27] S.T. Schindera, L. Diedrichsen, H.C. Müller, et al., Iterative reconstruction algorithm for abdominal multidetector CT at different tube voltages: assessment of diagnostic accuracy image quality, and radiation dose in a phantom study, Radiology 2 (2011) 454–462. [28] F. Holmquist, K. Hansson, F. Pasquariello, J. Björk, U. Nyman, Minimizing contrast medium doses to diagnose pulmonary embolism with 80-kvp multidetector computed tomography in azotemic patients, Acta Radiol. 2 (2009) 181–193. [29] L.S. Guimarães, J.G. Fletcher, W.S. Harmsen, et al., Appropriate patient selection at abdominal dual-energy CT using 80 kV: relationship between patient size image noise, and image quality, Radiology 3 (2010) 732–742. [30] Z. Szucs-Farkas, C. Schaller, S. Bensler, M.A. Patak, P. Vock, S.T. Schindera, Detection of pulmonary emboli with ct angiography at reduced radiation exposure and contrast material volume: comparison of 80 kVp and 120 kVp protocols in a matched cohort, Invest. Radiol. 12 (2009) 793–799. [31] Z. Szucs-Farkas, T. Strautz, M.A. Patak, L. Kurmann, P. Vock, S.T. Schindera, Is body weight the most appropriate criterion to select patients eligible for low-dose pulmonary CT angiography? Analysis of objective and subjective image quality at 80 kVp in 100 patients, Eur. Radiol. 8 (2009) 1914–1922. [32] A.R. Seyal, A. Arslanoglu, S.F. Abboud, A. Sahin, J.M. Horowitz, V. Yaghmai, CT of the abdomen with reduced tube voltage in adults: a practical approach, Radiographics 7 (2015) 1922–1939.