- Email: [email protected]
Radiation dose reduction in soft tissue neck CT using adaptive statistical iterative reconstruction (ASIR)
European Journal of Radiology 82 (2013) 2222–2226
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Radiation dose reduction in soft tissue neck CT using adaptive statistical iterative reconstruction (ASIR) Behroze Vachha a,∗ , Harald Brodoefel b , Carol Wilcox b , David B. Hackney b , Gul Moonis b a b
Neuroradiology Division, Massachusetts General Hospital, Harvard Medical School, 55 Fruit Street, Boston, MA 02114, United States Department of Radiology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, Boston, MA 02215, United States
a r t i c l e
i n f o
Article history: Received 9 February 2013 Received in revised form 30 July 2013 Accepted 4 August 2013 Keywords: Radiation-dose ASIR Neck CT
a b s t r a c t Purpose: To compare objective and subjective image quality in neck CT images acquired at different tube current–time products (275 mA s and 340 mA s) and reconstructed with filtered-back-projection (FBP) and adaptive statistical iterative reconstruction (ASIR). Materials and methods: HIPAA-compliant study with IRB approval and waiver of informed consent. 66 consecutive patients were randomly assigned to undergo contrast-enhanced neck CT at a standard tubecurrent–time-product (340 mA s; n = 33) or reduced tube-current–time-product (275 mA s, n = 33). Data sets were reconstructed with FBP and 2 levels (30%, 40%) of ASIR-FBP blending at 340 mA s and 275 mA s. Two neuroradiologists assessed subjective image quality in a blinded and randomized manner. Volume CT dose index (CTDIvol), dose-length-product (DLP), effective dose, and objective image noise were recorded. Signal-to-noise ratio (SNR) was computed as mean attenuation in a region of interest in the sternocleidomastoid muscle divided by image noise. Results: Compared with FBP, ASIR resulted in a reduction of image noise at both 340 mA s and 275 mA s. Reduction of tube current from 340 mA s to 275 mA s resulted in an increase in mean objective image noise (p = 0.02) and a decrease in SNR (p = 0.03) when images were reconstructed with FBP. However, when the 275 mA s images were reconstructed using ASIR, the mean objective image noise and SNR were similar to those of the standard 340 mA s CT images reconstructed with FBP (p > 0.05). Subjective image noise was ranked by both raters as either average or less-than-average irrespective of the tube current and iterative reconstruction technique. Conclusion: Adapting ASIR into neck CT protocols reduced effective dose by 17% without compromising image quality. © 2013 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Significant improvements in image quality, acquisition speed and patient throughput have led to a dramatic increase in the use of CT as an essential diagnostic tool for multiple clinical applications. As a result, there is growing concern regarding the accompanying radiation exposure and potential radiation-induced malignancies [1–4]. The “linear no-threshold” model accepted by many authors for stochastic effects posits a direct dose–response relationship between the development of solid cancers and exposure even to low doses of radiation [1]. Multiple dose-reducing strategies (X-ray beam filtration and collimation, manual tube current modulation tailored to patient size and indication, peak kilovoltage optimization, improved detector efficiency and noise reduction algorithms) are now included in newer CT scanners in an effort
∗ Corresponding author. Tel.: +1 617 726 8323; fax: +1 617 724 3338. E-mail address: [email protected] (B. Vachha). 0720-048X/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2013.08.014
to reduce radiation doses and hence the potential for radiationinduced malignancies. A technique to lower radiation dose, the reduction of tube current, is associated with unacceptable increases in image noise when the current standard reconstruction method of filtered back projection (FBP) is used to create the CT images. Iterative reconstruction (IR), routinely used for positron emission tomography (PET) and single photon emission computed tomography (SPECT), has been recently re-introduced to CT as an alternative mathematic algorithm that results in lower image noise than FBP [5–7]. The need for expensive computational hardware and increased reconstruction time prevented the implementation of IR for clinical use until recently. Adaptive Statistical Iterative Reconstruction (ASIR; GE Healthcare, Milwaukee, WI) is a recently developed IR algorithm approved by the US Food and Drug Administration for clinical use. ASIR reduces reconstruction time by using information obtained from the FBP algorithm as a starting point for image reconstruction and then repeatedly compares the estimated pixel value to the ideal value predicted by the noise model, until the estimated and ideal values converge [6].
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
Clinical studies supporting the use of iterative reconstruction methods such as ASIR in reducing radiation dose in head [8,9], temporal bone [10], cardiac [11–15], chest [16–18] and abdomen [19,20] have been reported. To our knowledge, the role of ASIR in reducing radiation dose in soft-tissue neck CT has not been reported. The present study compares objective and subjective image quality in soft-tissue neck CT images acquired at different tube current–time products (275 mA s and 340 mA s) and reconstructed with FBP and ASIR.
2223
levels. Each data set was coded, patient information was removed and the sets were randomized by a study co-author. 2.3. Dose measurements CT dose index volume (CTDIvol) and dose-length-product (DLP) were recorded for every examination. The effective dose in millisieverts (mSv) was estimated by multiplying the DLP with the region-specific conversion coefficient k (neck: 0.0052 mSv/(mGy cm)) [21–23].
2. Materials and methods
2.4. Objective measurements
2.1. Patient
Attenuation was measured in Hounsfield Units in a circular region of interest (ROI) in the sternocleidomastoid muscle at the level of the hyoid bone (ROI size 20–30 mm2 ). Objective image noise was defined as the standard deviation of attenuation measured in the air lateral to the cervical soft tissues at the level of the hyoid bone (ROI size 50–60 mm2 ). Signal-to-noise ratio (SNR) was calculated as mean attenuation of the sternocleidomastoid muscle at the level of the hyoid bone divided by objective image noise.
Our prospective clinical study was compliant with the Health Insurance Portability and Accountability Act (HIPAA) guidelines and was approved by the Human Research Committee of our Institutional Review Board (IRB). Informed written consent was waived because all studies were clinically indicated and performed as standard-of-care with a standard or lower-dose radiation protocol, and because early experience in our department showed that there was preservation of diagnostic image quality at all dose levels used in our study. Exclusion criteria for contrast-enhanced neck CT were as follows: age less than 18 years, impaired renal function (eGFR < 30 ml/min), hypersensitivity to iodine-containing contrast agents, and pregnancy. Sixty-six consecutive patients who were scheduled for contrastenhanced neck CTs for a variety of clinical conditions (predominantly initial evaluation or follow up of head and neck cancers and to rule out infectious/inflammatory pathology) were randomly assigned to standard dose (tube current–time product 340 mA s) or lower-dose (tube current–time product 275 mA s) protocols. Patient height and weight were noted at the time of the CT examination and body mass index (BMI) was computed.
2.2. Scanning technique and image reconstruction Clinically indicated standard-of-care neck CT examinations were performed with a 64-row multidetector commercial CT scanner (Discovery CT750 HD; GE Healthcare) with administration of an intravenous contrast agent (70 ml, Omnipaque 350; GE Healthcare, Princeton NJ). An initial pilot study in our department showed that a decrease in tube current–time product (mA s) by approximately 20% from the standard mA s implemented at our institution when combined with ASIR blending of 30–40% yielded diagnostically acceptable image quality (pilot data were not included as part of this study). The current study was designed based on these pilot analyses. We initiated this study to systematically evaluate if decrease in tube current–time product combined with iterative reconstruction using ASIR blending of 30% and 40% would be feasible clinically for neck CT examinations in comparison with the standard protocol using a larger sample size. Two different tube current–time products (340 mA s and 275 mA s) were chosen to represent standard and lower radiation doses. All other scanning parameters were held constant (tube voltage 120 kVp; pitch 0.984:1; table speed 39.34 mm per gantry rotation; helical acquisition mode; gantry rotation time 0.5 s; reconstructed section thickness 2.5 mm; reconstructed section interval 2.5 mm). Raw data of all CT examinations were reconstructed using FBP and ASIR techniques. ASIR was implemented using 2 different levels of blending (30% and 40%, hereafter referred throughout the text as ASIR30 and ASIR40 , respectively) at each of the two radiation dose
2.5. Subjective quality of FBP and ASIR images All neck CT image data sets were randomized and reviewed on a picture archiving and communication system diagnostic workstation (GE PACS, Centricity V 3.1.1.2, GE Healthcare, Milwaukee, WI) for assessment of subjective quality. Two neuroradiologists (R1 and R2, with 12 and 3 years experience, respectively) assessed all image data sets for image quality. Both raters were blinded to all clinical information and scanning parameters. We used image quality characteristics described by Singh et al. [18,20] as detailed below. Subjective image noise was assessed in soft tissue window settings (window width 450 HU; window level 50 HU) and graded on a five-point scale (1 = minimal image noise, 2 = less than average noise, 3 = average noise, 4 = above average noise, 5 = unacceptable noise). Artifacts were graded on a four-point scale (1 = no artifacts, 2 = minor artifacts not interfering with diagnosis, 3 = major artifacts affecting visualization of major structures, diagnosis still possible, 4 = artifacts affecting diagnosis). The main artifacts assessed were metallic or streak artifacts, pixilated appearance and truncation artifacts attributable to large body size. Visibility of small structures (i.e. small blood vessels, small lymph nodes < 1 cm and arytenoid cartilage) was graded on a five-point scale (1 = excellent visibility, 2 = above average visibility, 3 = average visibility, 4 = suboptimal visibility, 5 = unacceptable visibility). Visibility of small structures were assessed in soft tissue window settings (window width 450 HU; window level 50 HU) for small blood vessels and small lymph nodes < 1 cm and in bone windows (window width 3077 HU; window level 570 HU) for arytenoid cartilages. Neck pathology was documented as 1 = present and 2 = absent. If present, type of pathology was documented by each reviewer (e.g. tonsillitis, abscess, lymphadenopathy, tumor). This was verified for diagnostic accuracy by one of the authors separate from the reviewers and confirmed by correlating with diagnoses documented in medical records. Overall diagnostic confidence was assessed on a four-point scale (1 = completely confident in diagnosis, 2 = probably confident, 3 = confident only for a limited clinical situation such as a large lesion, 4 = not confident). 2.6. Statistical analysis Data were analyzed by using analysis of variance, independent samples and paired t-tests for objective metrics such as objective image noise, SNR, CT numbers and effective dose. Gender differences between the two groups were analyzed using the 2 test.
2224
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
Table 1 Patient characteristics and radiation dose in CT protocols with various tube currents. Characteristics
340 mA s (n = 33)
275 mA s (n = 33)
p
Age (years) Gender (male/female) BMI CTDIvol (mGy) DLP (mGy cm) Effective dose (mSv)
47 ± 16.14 (20.00–77.00) 16/17 29.19 ± 6.70 (19–50.38) 13.62 ± 0.09 (13.22–13.75) 388.53 ± 29.02 (317.90–462.54) 2.02 ± 0.15 (1.65–2.41)
52.4 ± 16.84 (21.00–88.00) 18/15 26.19 ± 6.67 (18.13–50) 11.16 ± 0.76 (10.87–15.39) 322.39 ± 43.06 (249.54–444.43) 1.68 ± 0.22 (1.30–2.31)
0.19 0.622 0.07 <0.001 <0.001 <0.001
Note: data (other than gender) are mean ± standard deviation, with ranges in parentheses.
Subjective image quality was assessed with the nonparametric Mann–Whitney U test. Interobserver variability was estimated by using both Cohen statistics and percentage agreement between the two radiologists for each of the assessed subjective image quality parameters. Definitions of levels of agreement on the basis of values were as follows: ≤ 0.20 indicated slight agreement; = 0.21–0.40, fair agreement; = 0.41–0.60, moderate agreement; = 0.61–0.80, substantial agreement; = 0.81–1, almost perfect agreement [24]. A value of p < 0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using SPSS statistical software (release 20.0 for Windows; SPSS, Chicago, IL). 3. Results The characteristics of subjects within the distinct study groups are shown in Table 1. There was no significant difference in BMI between the patients scanned with the standard dose (340 mA s) and the lower-dose (275 mA s) protocols, p = 0.07. 3.1. Radiation dose Compared with 2.02 mSv in the standard-dose group, the effective dose was reduced by 17% to 1.68 mSv in the lower tube current group (p < 0.001; Table 1). 3.2. Objective image quality Compared with FBP, ASIR resulted in reduction of objective image noise at both 340 mA s and 275 mA s. At 340 mA s, mean objective image noise decreased by 12% (p = 0.06) for ASIR30 images and by 15% (p = 0.03) for ASIR40 images (Table 2 and Fig. 1). At 275 mA s, mean objective image noise decreased by 25% (p = 0.009) for ASIR30 images and by 30% (p = 0.002) for ASIR40 images (Fig. 1). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a 25% increase in mean objective image noise (p = 0.02) when reconstructed with FBP (Fig. 1). However, when the 275 mA s images were reconstructed using ASIR there was no significant difference in mean objective image noise compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.27 for ASIR30 and p = 0.07 for ASIR40 image reconstructions). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a decrease in mean SNR (p = 0.03) when
reconstructed with FBP (Table 2). However, when the 275 mA s images were reconstructed using ASIR the resulting SNR was similar to the standard 340 mA s CT images reconstructed with FBP (p = 0.24 for ASIR30 and p = 0.19 for ASIR40 image reconstructions; Table 2). 3.3. Subjective image quality Overall, the interobserver agreement between the two radiologists was substantial for noise ( = 0.73), moderate for visibility of small structures ( = 0.47), slight for artifacts ( = 0.19) and slight for diagnostic confidence ( = 0.19). Percentage agreement between the two radiologists is summarized in Table 3. Neither the FBP nor the ASIR images were ranked as having unacceptable noise. At 340 mA s, use of an iterative algorithm resulted in a decrease in subjective image noise at ASIR30 (p = 0.05) and ASIR40 (p = 0.001) images when compared to images reconstructed with FBP. Similarly at 275 mA s, there was a decrease in subjective image noise in ASIR30 (p = 0.009) and ASIR40 (p < 0.001) images when compared to images reconstructed with FBP. Reduction of tube current–time products from 340 mA s to 275 mA s did not result in an increase in subjective image noise (p = 0.16) when these images were reconstructed using the FBP protocol. There was also no significant difference in subjective image noise when the 275 mA s images reconstructed using 30% ASIR-FBP blending (i.e. images at ASIR30 ) were compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.44). However, the 275 mA s images reconstructed using 40% ASIR-FBP blending (i.e. images at ASIR40 ) demonstrated significant decrease in subjective image noise when compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.02).
Table 2 Signal-to-noise ratio (SNR) in 66 contrast-enhanced neck CT studies. Reconstruction techniques
FBP ASIR30 ASIR40
Signal-to-noise ratio (SNR)
340 mA s (n = 33)
275 mA s (n = 33)
12.98 ± 5.30 (5.79–26.10) 14.67 ± 6.38 (2.25–28.93) 16.06 ± 6.99 (5.79–32.09)
10.56 ± 5.32 (2.77–26.59) 14.09 ± 7.49 (4.21–32.32) 14.15 ± 5.56 (5.51–29.90)
Note: data are mean signal to noise ratio (SNR) ± standard deviation, with ranges in parentheses.
Fig. 1. Objective image noise in contrast-enhanced soft tissue neck CT examinations.*p < 0.05.
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
2225
Table 3 Percentage agreement between the two radiologists for each of the assessed subjective image quality parameters. Reconstruction technique and tube current–time product (mA s) FBP 340 275 ASIR30 340 275 ASIR40 340 275
Noise
Visibility of small structures
Artifacts
Confidence in diagnosis
90.9 (30/33) 84.8 (28/33)
93.9 (31/33) 51.5 (17/33)
78.8 (26/33) 93.9 (31/33)
87.9 (29/33) 24.2 (8/33)
87.9 (29/33) 81.8 (27/33)
63.6 (21/33) 51.5 (17/33)
81.8 (27/33) 93.9 (31/33)
81.8 (27/33) 39.4 (13/33)
81.8 (27/33) 69.7 (23/33)
72.7 (24/33) 57.5 (19/33)
75.8 (25/33) 81.8 (27/33)
84.8 (28/33) 51.5 (17/33)
Note: data are percentage agreement between the two radiologists with raw data in parentheses.
Table 4 Mean ± standard deviation of qualitative image scores in 66 contrast-enhanced neck CT studies. Parameters
Tube current 340 mA s (n = 33) FBP
Image noise Visibility of small structures Artifacts Confidence in diagnosis
2.64 2.21 1.95 1.38
Tube current 275 mA s (n = 33)
ASIR30 ± ± ± ±
0.73 0.92 0.23 0.53
2.26 2.00 1.88 1.21
± ± ± ±
ASIR40 0.65 0.70 0.31 0.35
At 340 mA s, use of an iterative algorithm resulted in a decrease in visibility of small structures compared to FBP but these were not statistically significant (ASIR30 : p = 0.35; ASIR40 : p = 0.29). Similarly, at 275 mA s, use of an iterative algorithm resulted in a decrease in visibility of small structures which was not statistically significant for ASIR30 images (p = 0.18), but resulted in a statistically significant decrease in visibility of small structures at ASIR40 (p = 0.01). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a slight decrease in terms of visibility of small structures when these images were reconstructed with FBP, which was not statistically different (p = 0.17). There was also no significant difference in visibility of small structures when the 275 mA s images reconstructed using 30% ASIR-FBP blending (i.e. images at ASIR30 ) were compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.22). However, the 275 mA s images reconstructed using 40% ASIR-FBP blending (i.e. images at ASIR40 ) demonstrated significant decrease in visibility of small structures when compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.04). No significant image artifacts other than dental-amalgamrelated streak artifacts were seen on FBP, ASIR30 and ASIR40 images, regardless of the tube current–time product (p > 0.05). These were rated by both reviewers as minor artifacts that did not interfere with diagnosis. Both reviewers ranked confidence in diagnoses as either fully confident or probably confident. There were no significant differences in diagnostic confidence for FBP, ASIR30 and ASIR40 images regardless of the tube current–time product (p > 0.05). The percentage agreement between the radiologists with both FBP and ASIR for diagnostic acceptability of images was high at 340 mA s (81.8–87.9%) but decreased at 275 mA s (24.2–51.5%) irrespective of reconstruction technique (Table 3). Where neck abnormalities were present (in 20/33 patients imaged at 340 mA s and 18/33 patients imaged at 275 mA s), they were detected by both raters on all image sets with no effect of radiation dose or reconstruction method. Table 4 provides the means and standard deviations for the subjective image quality data. 4. Discussion Neck CT is the imaging modality of choice to evaluate neck pathology. A neck CT exam extends from the posterior fossa to the
1.94 1.97 1.91 1.23
± ± ± ±
FBP 0.78 0.70 0.34 0.38
2.83 2.08 1.97 1.45
ASIR30 ± ± ± ±
0.51 0.50 0.12 0.36
2.47 1.94 1.94 1.41
± ± ± ±
ASIR40 0.53 0.50 0.21 0.40
2.23 1.76 1.94 1.36
± ± ± ±
0.56 0.47 0.21 0.40
aortic arch, irradiating radiosensitive organs such as the eyes and the thyroid, which have higher stochastic risk for injury and future malignancy, particularly with cumulative radiation exposure [3]. Several studies have already confirmed the capability of ASIR to achieve robust image quality in low dose CT examinations of the head, temporal bone, thorax and abdomen [8,16–18,20]. To the best of our knowledge, this study is the first to assess an iterative reconstruction approach in soft tissue neck CT examinations. FBP techniques of image reconstruction result in high image noise at low radiation doses [5]. Unlike traditional FBP, ASIR is based on a correction loop within the image-generation process, which leads to significant reduction of noise independent of the dose. The advantage of ASIR is most apparent in low dose acquisitions where noise obscures visualization of clinically relevant information. Our study demonstrated a decrease in objective and subjective image noise in ASIR images compared to FBP images regardless of the tube current–time products. Lower objective noise levels using ASIR resulted in increased SNR. In addition, by adapting ASIR into our protocols, we were able to reduce the tube current–time product from 340 mA s to 275 mA s without compromising objective and subjective image quality. Overall images reconstructed with FBP demonstrated slightly better visibility of small structures compared to those reconstructed with ASIR, however these differences were statistically significant and more pronounced only for the 275 mA s images reconstructed with 40% ASIR blending. We attribute this decrease in sharpness and spatial resolution to the increased “smoothening” effect of ASIR at 40% ASIR-FBP blending. This trade-off between higher SNR at ASIR40 and loss of fine detail needs to be further evaluated in a larger series. Despite this, where pathology was present, it was always detected by both reviewers with a 100% diagnostic accuracy irrespective of tube current and reconstruction technique used. In our study overall diagnostic acceptability of all image sets were not affected irrespective of tube current and iterative reconstruction method and no images were ranked as being diagnostically unacceptable. It was noted, however, that the percentage agreement between the radiologists with both FBP and ASIR for diagnostic acceptability of images was high at 340 mA s but decreased at 275 mA s irrespective of reconstruction technique. This may be related to experience such that the radiologist with 12 years experience ranked diagnostic confidence mainly as “completely confident”
2226
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
with both the standard and lower-dose protocols, while the less experienced radiologist ranked diagnostic confidence mainly as “probably confident” in diagnosis when faced with the lower-dose images. There are limitations to our study – the small sample size per group requires confirmation of our findings in a larger patient series. Different tube currents were applied to different patient cohorts and a more accurate intra-patient comparison of protocols was not performed since that would require scanning the patients twice in the same session, and few patients undergo repeat imaging in a short time period. Evaluating patients with neoplasms who undergo serial follow up examinations with CT may address the issue of intra-subject comparison, however accurate comparisons of lesion conspicuity or size may be confounded by treatment effects and time elapsed between studies. Another limitation is that ASIR images appear “smoothed” compared to the FBP images that radiologists have become accustomed to [20]. It is possible, therefore, that these differences in image appearance may have allowed the two reviewers to distinguish between ASIR and FBP images despite de-identification and randomization of image sets. As Singh et al. [20] have discussed, there is also a concern that repetitive reviewing of image data sets (FBP, ASIR30 and ASIR40 in each patient for both dose levels) may have aided the radiologist in detecting pathology. To minimize this bias, we used a methodology similar to Singh et al. [20] in which the radiologists were asked to first evaluate the images with the most objective noise and subsequently evaluate image sets with decreasing levels of objective noise. Finally, we did not evaluate the effect of the low-dose protocol on non-contrast neck CT examinations, which may be less acceptable clinically than the contrast-enhanced CT examinations. The practical application of our study is a 17% reduction in effective dose in soft tissue neck CT examinations when using ASIR without compromising image quality or diagnostic acceptability. Because all the examinations in the lower mA s group were considered completely or probably acceptable for diagnosis, we have incorporated the lower dose settings with FBP and ASIR as our routine contrast enhanced soft tissue neck CT protocol. The absence of detectable subjective differences in image quality between the 340 mA s and 275 mA s group even when reconstructed with standard FBP, suggests that we may be able to reduce the tube current–time product even further for both FBP and ASIR reconstructions, with the data suggesting that the potential dose reduction could be higher for ASIR than for FBP. Additionally, in future studies, we plan on incrementally increasing ASIR blending to evaluate the highest ASIR blending percentage at which there is a loss of diagnostic imaging quality. Combining ASIR techniques with other dose-reduction strategies such as automatic tube current modulation would conceivably enable even further dose reduction and should be evaluated in a larger series. As model-based iterative reconstruction techniques evolve, further reductions in radiation dose may be possible, and this process of evaluating image noise and quality will need to be repeated at each step. Conflicts of interest None of the authors have any conflicts of interest.
Acknowledgement The authors thank Dr. J.A. Parker for statistical guidance. References [1] Radiation CtAHRfEtLLoI, Council NR. Health risks from exposure to low levels of ionizing radiation: BEIR VII phase 2. The National Academies Press; 2006. [2] Amis Jr ES, Butler PF, Applegate KE, et al. American College of Radiology white paper on radiation dose in medicine. J Am Coll Radiol 2007;4(5): 272–84. [3] Brenner DJ, Hall EJ. Computed tomography – an increasing source of radiation exposure. N Engl J Med 2007;357(22):2277–84. [4] Lauer MS. Elements of danger – the case of medical imaging. N Engl J Med 2009;361(9):841–3. [5] Hara AK, Paden RG, Silva AC, Kujak JL, Lawder HJ, Pavlicek W. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. AJR Am J Roentgenol 2009;193(3):764–71. [6] Silva AC, Lawder HJ, Hara A, Kujak J, Pavlicek W. Innovations in CT dose reduction strategy: application of the adaptive statistical iterative reconstruction algorithm. AJR Am J Roentgenol 2010;194(1):191–9. [7] Thibault JB, Sauer KD, Bouman CA, Hsieh J. A three-dimensional statistical approach to improved image quality for multislice helical CT. Med Phys 2007;34(11):4526–44. [8] Kilic K, Erbas G, Guryildirim M, Arac M, Ilgit E, Coskun B. Lowering the dose in head CT using adaptive statistical iterative reconstruction. AJNR Am J Neuroradiol 2011;32(9):1578–82. [9] Korn A, Fenchel M, Bender B, et al. Iterative reconstruction in head CT: image quality of routine and low-dose protocols in comparison with standard filtered back-projection. AJNR Am J Neuroradiol 2012;33(2):218–24. [10] Niu Y, Wang Z, Liu Y, Liu Z, Yao V. Radiation dose to the lens using different temporal bone CT scanning protocols. AJNR Am J Neuroradiol 2010;31(2):226–9. [11] Tumur O, Soon K, Brown F, Mykytowycz M. New scanning technique using adaptive statistical iterative reconstruction (ASIR) significantly reduced the radiation dose of cardiac CT. J Med Imaging Radiat Oncol 2013;57(3):292–6. [12] Gebhard C, Fuchs TA, Fiechter M, et al. Image quality of low-dose CCTA in obese patients: impact of high-definition computed tomography and adaptive statistical iterative reconstruction. Int J Cardiovasc Imaging 2013 [Epub ahead of print]. [13] Gebhard C, Fiechter M, Fuchs TA, et al. Coronary artery stents: influence of adaptive statistical iterative reconstruction on image quality using 64-HDCT. Eur Heart J Cardiovasc Imaging 2013 [Epub ahead of print]. [14] Fuchs TA, Fiechter M, Gebhard C, et al. CT coronary angiography: impact of adapted statistical iterative reconstruction (ASIR) on coronary stenosis and plaque composition analysis. Int J Cardiovasc Imaging 2013;29(3):719–24. [15] Gebhard C, Fiechter M, Fuchs TA, et al. Coronary artery calcium scoring: influence of adaptive statistical iterative reconstruction using 64-MDCT. Int J Cardiol 2012 [Epub ahead of print]. [16] Pontana F, Duhamel A, Pagniez J, et al. Chest computed tomography using iterative reconstruction vs filtered back projection (Part 2): image quality of low-dose CT examinations in 80 patients. Eur Radiol 2011;21(3):636–43. [17] Pontana F, Pagniez J, Flohr T, et al. Chest computed tomography using iterative reconstruction vs filtered back projection (Part 1): evaluation of image noise reduction in 32 patients. Eur Radiol 2011;21(3):627–35. [18] Singh S, Kalra MK, Gilman MD, et al. Adaptive statistical iterative reconstruction technique for radiation dose reduction in chest CT: a pilot study. Radiology 2011;259(2):565–73. [19] Shuman WP, Green DE, Busey JM, et al. Model-based iterative reconstruction versus adaptive statistical iterative reconstruction and filtered back projection in liver 64-MDCT: focal lesion detection, lesion conspicuity, and image noise. AJR Am J Roentgenol 2013;200(5):1071–6. [20] Singh S, Kalra MK, Hsieh J, et al. Abdominal CT: comparison of adaptive statistical iterative and filtered back projection reconstruction techniques. Radiology 2010;257(2):373–83. [21] The 2007 Recommendations of the International Commission on Radiological Protection. ICRP publication 103. Ann ICRP 2007;37(2–4):1–332. [22] Deak PD, Smal Y, Kalender WA. Multisection CT protocols: sex- and age-specific conversion factors used to determine effective dose from dose-length product. Radiology 2010;257(1):158–66. [23] Kalender W. Computed tomography: fundamentals, system technology, image quality, applications. Erlangen: Publicis Publishing; 2011. [24] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–74.
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Radiation dose reduction in soft tissue neck CT using adaptive statistical iterative reconstruction (ASIR) Behroze Vachha a,∗ , Harald Brodoefel b , Carol Wilcox b , David B. Hackney b , Gul Moonis b a b
Neuroradiology Division, Massachusetts General Hospital, Harvard Medical School, 55 Fruit Street, Boston, MA 02114, United States Department of Radiology, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, Boston, MA 02215, United States
a r t i c l e
i n f o
Article history: Received 9 February 2013 Received in revised form 30 July 2013 Accepted 4 August 2013 Keywords: Radiation-dose ASIR Neck CT
a b s t r a c t Purpose: To compare objective and subjective image quality in neck CT images acquired at different tube current–time products (275 mA s and 340 mA s) and reconstructed with filtered-back-projection (FBP) and adaptive statistical iterative reconstruction (ASIR). Materials and methods: HIPAA-compliant study with IRB approval and waiver of informed consent. 66 consecutive patients were randomly assigned to undergo contrast-enhanced neck CT at a standard tubecurrent–time-product (340 mA s; n = 33) or reduced tube-current–time-product (275 mA s, n = 33). Data sets were reconstructed with FBP and 2 levels (30%, 40%) of ASIR-FBP blending at 340 mA s and 275 mA s. Two neuroradiologists assessed subjective image quality in a blinded and randomized manner. Volume CT dose index (CTDIvol), dose-length-product (DLP), effective dose, and objective image noise were recorded. Signal-to-noise ratio (SNR) was computed as mean attenuation in a region of interest in the sternocleidomastoid muscle divided by image noise. Results: Compared with FBP, ASIR resulted in a reduction of image noise at both 340 mA s and 275 mA s. Reduction of tube current from 340 mA s to 275 mA s resulted in an increase in mean objective image noise (p = 0.02) and a decrease in SNR (p = 0.03) when images were reconstructed with FBP. However, when the 275 mA s images were reconstructed using ASIR, the mean objective image noise and SNR were similar to those of the standard 340 mA s CT images reconstructed with FBP (p > 0.05). Subjective image noise was ranked by both raters as either average or less-than-average irrespective of the tube current and iterative reconstruction technique. Conclusion: Adapting ASIR into neck CT protocols reduced effective dose by 17% without compromising image quality. © 2013 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Significant improvements in image quality, acquisition speed and patient throughput have led to a dramatic increase in the use of CT as an essential diagnostic tool for multiple clinical applications. As a result, there is growing concern regarding the accompanying radiation exposure and potential radiation-induced malignancies [1–4]. The “linear no-threshold” model accepted by many authors for stochastic effects posits a direct dose–response relationship between the development of solid cancers and exposure even to low doses of radiation [1]. Multiple dose-reducing strategies (X-ray beam filtration and collimation, manual tube current modulation tailored to patient size and indication, peak kilovoltage optimization, improved detector efficiency and noise reduction algorithms) are now included in newer CT scanners in an effort
∗ Corresponding author. Tel.: +1 617 726 8323; fax: +1 617 724 3338. E-mail address: [email protected] (B. Vachha). 0720-048X/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2013.08.014
to reduce radiation doses and hence the potential for radiationinduced malignancies. A technique to lower radiation dose, the reduction of tube current, is associated with unacceptable increases in image noise when the current standard reconstruction method of filtered back projection (FBP) is used to create the CT images. Iterative reconstruction (IR), routinely used for positron emission tomography (PET) and single photon emission computed tomography (SPECT), has been recently re-introduced to CT as an alternative mathematic algorithm that results in lower image noise than FBP [5–7]. The need for expensive computational hardware and increased reconstruction time prevented the implementation of IR for clinical use until recently. Adaptive Statistical Iterative Reconstruction (ASIR; GE Healthcare, Milwaukee, WI) is a recently developed IR algorithm approved by the US Food and Drug Administration for clinical use. ASIR reduces reconstruction time by using information obtained from the FBP algorithm as a starting point for image reconstruction and then repeatedly compares the estimated pixel value to the ideal value predicted by the noise model, until the estimated and ideal values converge [6].
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
Clinical studies supporting the use of iterative reconstruction methods such as ASIR in reducing radiation dose in head [8,9], temporal bone [10], cardiac [11–15], chest [16–18] and abdomen [19,20] have been reported. To our knowledge, the role of ASIR in reducing radiation dose in soft-tissue neck CT has not been reported. The present study compares objective and subjective image quality in soft-tissue neck CT images acquired at different tube current–time products (275 mA s and 340 mA s) and reconstructed with FBP and ASIR.
2223
levels. Each data set was coded, patient information was removed and the sets were randomized by a study co-author. 2.3. Dose measurements CT dose index volume (CTDIvol) and dose-length-product (DLP) were recorded for every examination. The effective dose in millisieverts (mSv) was estimated by multiplying the DLP with the region-specific conversion coefficient k (neck: 0.0052 mSv/(mGy cm)) [21–23].
2. Materials and methods
2.4. Objective measurements
2.1. Patient
Attenuation was measured in Hounsfield Units in a circular region of interest (ROI) in the sternocleidomastoid muscle at the level of the hyoid bone (ROI size 20–30 mm2 ). Objective image noise was defined as the standard deviation of attenuation measured in the air lateral to the cervical soft tissues at the level of the hyoid bone (ROI size 50–60 mm2 ). Signal-to-noise ratio (SNR) was calculated as mean attenuation of the sternocleidomastoid muscle at the level of the hyoid bone divided by objective image noise.
Our prospective clinical study was compliant with the Health Insurance Portability and Accountability Act (HIPAA) guidelines and was approved by the Human Research Committee of our Institutional Review Board (IRB). Informed written consent was waived because all studies were clinically indicated and performed as standard-of-care with a standard or lower-dose radiation protocol, and because early experience in our department showed that there was preservation of diagnostic image quality at all dose levels used in our study. Exclusion criteria for contrast-enhanced neck CT were as follows: age less than 18 years, impaired renal function (eGFR < 30 ml/min), hypersensitivity to iodine-containing contrast agents, and pregnancy. Sixty-six consecutive patients who were scheduled for contrastenhanced neck CTs for a variety of clinical conditions (predominantly initial evaluation or follow up of head and neck cancers and to rule out infectious/inflammatory pathology) were randomly assigned to standard dose (tube current–time product 340 mA s) or lower-dose (tube current–time product 275 mA s) protocols. Patient height and weight were noted at the time of the CT examination and body mass index (BMI) was computed.
2.2. Scanning technique and image reconstruction Clinically indicated standard-of-care neck CT examinations were performed with a 64-row multidetector commercial CT scanner (Discovery CT750 HD; GE Healthcare) with administration of an intravenous contrast agent (70 ml, Omnipaque 350; GE Healthcare, Princeton NJ). An initial pilot study in our department showed that a decrease in tube current–time product (mA s) by approximately 20% from the standard mA s implemented at our institution when combined with ASIR blending of 30–40% yielded diagnostically acceptable image quality (pilot data were not included as part of this study). The current study was designed based on these pilot analyses. We initiated this study to systematically evaluate if decrease in tube current–time product combined with iterative reconstruction using ASIR blending of 30% and 40% would be feasible clinically for neck CT examinations in comparison with the standard protocol using a larger sample size. Two different tube current–time products (340 mA s and 275 mA s) were chosen to represent standard and lower radiation doses. All other scanning parameters were held constant (tube voltage 120 kVp; pitch 0.984:1; table speed 39.34 mm per gantry rotation; helical acquisition mode; gantry rotation time 0.5 s; reconstructed section thickness 2.5 mm; reconstructed section interval 2.5 mm). Raw data of all CT examinations were reconstructed using FBP and ASIR techniques. ASIR was implemented using 2 different levels of blending (30% and 40%, hereafter referred throughout the text as ASIR30 and ASIR40 , respectively) at each of the two radiation dose
2.5. Subjective quality of FBP and ASIR images All neck CT image data sets were randomized and reviewed on a picture archiving and communication system diagnostic workstation (GE PACS, Centricity V 3.1.1.2, GE Healthcare, Milwaukee, WI) for assessment of subjective quality. Two neuroradiologists (R1 and R2, with 12 and 3 years experience, respectively) assessed all image data sets for image quality. Both raters were blinded to all clinical information and scanning parameters. We used image quality characteristics described by Singh et al. [18,20] as detailed below. Subjective image noise was assessed in soft tissue window settings (window width 450 HU; window level 50 HU) and graded on a five-point scale (1 = minimal image noise, 2 = less than average noise, 3 = average noise, 4 = above average noise, 5 = unacceptable noise). Artifacts were graded on a four-point scale (1 = no artifacts, 2 = minor artifacts not interfering with diagnosis, 3 = major artifacts affecting visualization of major structures, diagnosis still possible, 4 = artifacts affecting diagnosis). The main artifacts assessed were metallic or streak artifacts, pixilated appearance and truncation artifacts attributable to large body size. Visibility of small structures (i.e. small blood vessels, small lymph nodes < 1 cm and arytenoid cartilage) was graded on a five-point scale (1 = excellent visibility, 2 = above average visibility, 3 = average visibility, 4 = suboptimal visibility, 5 = unacceptable visibility). Visibility of small structures were assessed in soft tissue window settings (window width 450 HU; window level 50 HU) for small blood vessels and small lymph nodes < 1 cm and in bone windows (window width 3077 HU; window level 570 HU) for arytenoid cartilages. Neck pathology was documented as 1 = present and 2 = absent. If present, type of pathology was documented by each reviewer (e.g. tonsillitis, abscess, lymphadenopathy, tumor). This was verified for diagnostic accuracy by one of the authors separate from the reviewers and confirmed by correlating with diagnoses documented in medical records. Overall diagnostic confidence was assessed on a four-point scale (1 = completely confident in diagnosis, 2 = probably confident, 3 = confident only for a limited clinical situation such as a large lesion, 4 = not confident). 2.6. Statistical analysis Data were analyzed by using analysis of variance, independent samples and paired t-tests for objective metrics such as objective image noise, SNR, CT numbers and effective dose. Gender differences between the two groups were analyzed using the 2 test.
2224
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
Table 1 Patient characteristics and radiation dose in CT protocols with various tube currents. Characteristics
340 mA s (n = 33)
275 mA s (n = 33)
p
Age (years) Gender (male/female) BMI CTDIvol (mGy) DLP (mGy cm) Effective dose (mSv)
47 ± 16.14 (20.00–77.00) 16/17 29.19 ± 6.70 (19–50.38) 13.62 ± 0.09 (13.22–13.75) 388.53 ± 29.02 (317.90–462.54) 2.02 ± 0.15 (1.65–2.41)
52.4 ± 16.84 (21.00–88.00) 18/15 26.19 ± 6.67 (18.13–50) 11.16 ± 0.76 (10.87–15.39) 322.39 ± 43.06 (249.54–444.43) 1.68 ± 0.22 (1.30–2.31)
0.19 0.622 0.07 <0.001 <0.001 <0.001
Note: data (other than gender) are mean ± standard deviation, with ranges in parentheses.
Subjective image quality was assessed with the nonparametric Mann–Whitney U test. Interobserver variability was estimated by using both Cohen statistics and percentage agreement between the two radiologists for each of the assessed subjective image quality parameters. Definitions of levels of agreement on the basis of values were as follows: ≤ 0.20 indicated slight agreement; = 0.21–0.40, fair agreement; = 0.41–0.60, moderate agreement; = 0.61–0.80, substantial agreement; = 0.81–1, almost perfect agreement [24]. A value of p < 0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using SPSS statistical software (release 20.0 for Windows; SPSS, Chicago, IL). 3. Results The characteristics of subjects within the distinct study groups are shown in Table 1. There was no significant difference in BMI between the patients scanned with the standard dose (340 mA s) and the lower-dose (275 mA s) protocols, p = 0.07. 3.1. Radiation dose Compared with 2.02 mSv in the standard-dose group, the effective dose was reduced by 17% to 1.68 mSv in the lower tube current group (p < 0.001; Table 1). 3.2. Objective image quality Compared with FBP, ASIR resulted in reduction of objective image noise at both 340 mA s and 275 mA s. At 340 mA s, mean objective image noise decreased by 12% (p = 0.06) for ASIR30 images and by 15% (p = 0.03) for ASIR40 images (Table 2 and Fig. 1). At 275 mA s, mean objective image noise decreased by 25% (p = 0.009) for ASIR30 images and by 30% (p = 0.002) for ASIR40 images (Fig. 1). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a 25% increase in mean objective image noise (p = 0.02) when reconstructed with FBP (Fig. 1). However, when the 275 mA s images were reconstructed using ASIR there was no significant difference in mean objective image noise compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.27 for ASIR30 and p = 0.07 for ASIR40 image reconstructions). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a decrease in mean SNR (p = 0.03) when
reconstructed with FBP (Table 2). However, when the 275 mA s images were reconstructed using ASIR the resulting SNR was similar to the standard 340 mA s CT images reconstructed with FBP (p = 0.24 for ASIR30 and p = 0.19 for ASIR40 image reconstructions; Table 2). 3.3. Subjective image quality Overall, the interobserver agreement between the two radiologists was substantial for noise ( = 0.73), moderate for visibility of small structures ( = 0.47), slight for artifacts ( = 0.19) and slight for diagnostic confidence ( = 0.19). Percentage agreement between the two radiologists is summarized in Table 3. Neither the FBP nor the ASIR images were ranked as having unacceptable noise. At 340 mA s, use of an iterative algorithm resulted in a decrease in subjective image noise at ASIR30 (p = 0.05) and ASIR40 (p = 0.001) images when compared to images reconstructed with FBP. Similarly at 275 mA s, there was a decrease in subjective image noise in ASIR30 (p = 0.009) and ASIR40 (p < 0.001) images when compared to images reconstructed with FBP. Reduction of tube current–time products from 340 mA s to 275 mA s did not result in an increase in subjective image noise (p = 0.16) when these images were reconstructed using the FBP protocol. There was also no significant difference in subjective image noise when the 275 mA s images reconstructed using 30% ASIR-FBP blending (i.e. images at ASIR30 ) were compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.44). However, the 275 mA s images reconstructed using 40% ASIR-FBP blending (i.e. images at ASIR40 ) demonstrated significant decrease in subjective image noise when compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.02).
Table 2 Signal-to-noise ratio (SNR) in 66 contrast-enhanced neck CT studies. Reconstruction techniques
FBP ASIR30 ASIR40
Signal-to-noise ratio (SNR)
340 mA s (n = 33)
275 mA s (n = 33)
12.98 ± 5.30 (5.79–26.10) 14.67 ± 6.38 (2.25–28.93) 16.06 ± 6.99 (5.79–32.09)
10.56 ± 5.32 (2.77–26.59) 14.09 ± 7.49 (4.21–32.32) 14.15 ± 5.56 (5.51–29.90)
Note: data are mean signal to noise ratio (SNR) ± standard deviation, with ranges in parentheses.
Fig. 1. Objective image noise in contrast-enhanced soft tissue neck CT examinations.*p < 0.05.
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
2225
Table 3 Percentage agreement between the two radiologists for each of the assessed subjective image quality parameters. Reconstruction technique and tube current–time product (mA s) FBP 340 275 ASIR30 340 275 ASIR40 340 275
Noise
Visibility of small structures
Artifacts
Confidence in diagnosis
90.9 (30/33) 84.8 (28/33)
93.9 (31/33) 51.5 (17/33)
78.8 (26/33) 93.9 (31/33)
87.9 (29/33) 24.2 (8/33)
87.9 (29/33) 81.8 (27/33)
63.6 (21/33) 51.5 (17/33)
81.8 (27/33) 93.9 (31/33)
81.8 (27/33) 39.4 (13/33)
81.8 (27/33) 69.7 (23/33)
72.7 (24/33) 57.5 (19/33)
75.8 (25/33) 81.8 (27/33)
84.8 (28/33) 51.5 (17/33)
Note: data are percentage agreement between the two radiologists with raw data in parentheses.
Table 4 Mean ± standard deviation of qualitative image scores in 66 contrast-enhanced neck CT studies. Parameters
Tube current 340 mA s (n = 33) FBP
Image noise Visibility of small structures Artifacts Confidence in diagnosis
2.64 2.21 1.95 1.38
Tube current 275 mA s (n = 33)
ASIR30 ± ± ± ±
0.73 0.92 0.23 0.53
2.26 2.00 1.88 1.21
± ± ± ±
ASIR40 0.65 0.70 0.31 0.35
At 340 mA s, use of an iterative algorithm resulted in a decrease in visibility of small structures compared to FBP but these were not statistically significant (ASIR30 : p = 0.35; ASIR40 : p = 0.29). Similarly, at 275 mA s, use of an iterative algorithm resulted in a decrease in visibility of small structures which was not statistically significant for ASIR30 images (p = 0.18), but resulted in a statistically significant decrease in visibility of small structures at ASIR40 (p = 0.01). Reduction of tube current–time products from 340 mA s to 275 mA s resulted in a slight decrease in terms of visibility of small structures when these images were reconstructed with FBP, which was not statistically different (p = 0.17). There was also no significant difference in visibility of small structures when the 275 mA s images reconstructed using 30% ASIR-FBP blending (i.e. images at ASIR30 ) were compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.22). However, the 275 mA s images reconstructed using 40% ASIR-FBP blending (i.e. images at ASIR40 ) demonstrated significant decrease in visibility of small structures when compared to the standard 340 mA s CT images reconstructed with FBP (p = 0.04). No significant image artifacts other than dental-amalgamrelated streak artifacts were seen on FBP, ASIR30 and ASIR40 images, regardless of the tube current–time product (p > 0.05). These were rated by both reviewers as minor artifacts that did not interfere with diagnosis. Both reviewers ranked confidence in diagnoses as either fully confident or probably confident. There were no significant differences in diagnostic confidence for FBP, ASIR30 and ASIR40 images regardless of the tube current–time product (p > 0.05). The percentage agreement between the radiologists with both FBP and ASIR for diagnostic acceptability of images was high at 340 mA s (81.8–87.9%) but decreased at 275 mA s (24.2–51.5%) irrespective of reconstruction technique (Table 3). Where neck abnormalities were present (in 20/33 patients imaged at 340 mA s and 18/33 patients imaged at 275 mA s), they were detected by both raters on all image sets with no effect of radiation dose or reconstruction method. Table 4 provides the means and standard deviations for the subjective image quality data. 4. Discussion Neck CT is the imaging modality of choice to evaluate neck pathology. A neck CT exam extends from the posterior fossa to the
1.94 1.97 1.91 1.23
± ± ± ±
FBP 0.78 0.70 0.34 0.38
2.83 2.08 1.97 1.45
ASIR30 ± ± ± ±
0.51 0.50 0.12 0.36
2.47 1.94 1.94 1.41
± ± ± ±
ASIR40 0.53 0.50 0.21 0.40
2.23 1.76 1.94 1.36
± ± ± ±
0.56 0.47 0.21 0.40
aortic arch, irradiating radiosensitive organs such as the eyes and the thyroid, which have higher stochastic risk for injury and future malignancy, particularly with cumulative radiation exposure [3]. Several studies have already confirmed the capability of ASIR to achieve robust image quality in low dose CT examinations of the head, temporal bone, thorax and abdomen [8,16–18,20]. To the best of our knowledge, this study is the first to assess an iterative reconstruction approach in soft tissue neck CT examinations. FBP techniques of image reconstruction result in high image noise at low radiation doses [5]. Unlike traditional FBP, ASIR is based on a correction loop within the image-generation process, which leads to significant reduction of noise independent of the dose. The advantage of ASIR is most apparent in low dose acquisitions where noise obscures visualization of clinically relevant information. Our study demonstrated a decrease in objective and subjective image noise in ASIR images compared to FBP images regardless of the tube current–time products. Lower objective noise levels using ASIR resulted in increased SNR. In addition, by adapting ASIR into our protocols, we were able to reduce the tube current–time product from 340 mA s to 275 mA s without compromising objective and subjective image quality. Overall images reconstructed with FBP demonstrated slightly better visibility of small structures compared to those reconstructed with ASIR, however these differences were statistically significant and more pronounced only for the 275 mA s images reconstructed with 40% ASIR blending. We attribute this decrease in sharpness and spatial resolution to the increased “smoothening” effect of ASIR at 40% ASIR-FBP blending. This trade-off between higher SNR at ASIR40 and loss of fine detail needs to be further evaluated in a larger series. Despite this, where pathology was present, it was always detected by both reviewers with a 100% diagnostic accuracy irrespective of tube current and reconstruction technique used. In our study overall diagnostic acceptability of all image sets were not affected irrespective of tube current and iterative reconstruction method and no images were ranked as being diagnostically unacceptable. It was noted, however, that the percentage agreement between the radiologists with both FBP and ASIR for diagnostic acceptability of images was high at 340 mA s but decreased at 275 mA s irrespective of reconstruction technique. This may be related to experience such that the radiologist with 12 years experience ranked diagnostic confidence mainly as “completely confident”
2226
B. Vachha et al. / European Journal of Radiology 82 (2013) 2222–2226
with both the standard and lower-dose protocols, while the less experienced radiologist ranked diagnostic confidence mainly as “probably confident” in diagnosis when faced with the lower-dose images. There are limitations to our study – the small sample size per group requires confirmation of our findings in a larger patient series. Different tube currents were applied to different patient cohorts and a more accurate intra-patient comparison of protocols was not performed since that would require scanning the patients twice in the same session, and few patients undergo repeat imaging in a short time period. Evaluating patients with neoplasms who undergo serial follow up examinations with CT may address the issue of intra-subject comparison, however accurate comparisons of lesion conspicuity or size may be confounded by treatment effects and time elapsed between studies. Another limitation is that ASIR images appear “smoothed” compared to the FBP images that radiologists have become accustomed to [20]. It is possible, therefore, that these differences in image appearance may have allowed the two reviewers to distinguish between ASIR and FBP images despite de-identification and randomization of image sets. As Singh et al. [20] have discussed, there is also a concern that repetitive reviewing of image data sets (FBP, ASIR30 and ASIR40 in each patient for both dose levels) may have aided the radiologist in detecting pathology. To minimize this bias, we used a methodology similar to Singh et al. [20] in which the radiologists were asked to first evaluate the images with the most objective noise and subsequently evaluate image sets with decreasing levels of objective noise. Finally, we did not evaluate the effect of the low-dose protocol on non-contrast neck CT examinations, which may be less acceptable clinically than the contrast-enhanced CT examinations. The practical application of our study is a 17% reduction in effective dose in soft tissue neck CT examinations when using ASIR without compromising image quality or diagnostic acceptability. Because all the examinations in the lower mA s group were considered completely or probably acceptable for diagnosis, we have incorporated the lower dose settings with FBP and ASIR as our routine contrast enhanced soft tissue neck CT protocol. The absence of detectable subjective differences in image quality between the 340 mA s and 275 mA s group even when reconstructed with standard FBP, suggests that we may be able to reduce the tube current–time product even further for both FBP and ASIR reconstructions, with the data suggesting that the potential dose reduction could be higher for ASIR than for FBP. Additionally, in future studies, we plan on incrementally increasing ASIR blending to evaluate the highest ASIR blending percentage at which there is a loss of diagnostic imaging quality. Combining ASIR techniques with other dose-reduction strategies such as automatic tube current modulation would conceivably enable even further dose reduction and should be evaluated in a larger series. As model-based iterative reconstruction techniques evolve, further reductions in radiation dose may be possible, and this process of evaluating image noise and quality will need to be repeated at each step. Conflicts of interest None of the authors have any conflicts of interest.
Acknowledgement The authors thank Dr. J.A. Parker for statistical guidance. References [1] Radiation CtAHRfEtLLoI, Council NR. Health risks from exposure to low levels of ionizing radiation: BEIR VII phase 2. The National Academies Press; 2006. [2] Amis Jr ES, Butler PF, Applegate KE, et al. American College of Radiology white paper on radiation dose in medicine. J Am Coll Radiol 2007;4(5): 272–84. [3] Brenner DJ, Hall EJ. Computed tomography – an increasing source of radiation exposure. N Engl J Med 2007;357(22):2277–84. [4] Lauer MS. Elements of danger – the case of medical imaging. N Engl J Med 2009;361(9):841–3. [5] Hara AK, Paden RG, Silva AC, Kujak JL, Lawder HJ, Pavlicek W. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. AJR Am J Roentgenol 2009;193(3):764–71. [6] Silva AC, Lawder HJ, Hara A, Kujak J, Pavlicek W. Innovations in CT dose reduction strategy: application of the adaptive statistical iterative reconstruction algorithm. AJR Am J Roentgenol 2010;194(1):191–9. [7] Thibault JB, Sauer KD, Bouman CA, Hsieh J. A three-dimensional statistical approach to improved image quality for multislice helical CT. Med Phys 2007;34(11):4526–44. [8] Kilic K, Erbas G, Guryildirim M, Arac M, Ilgit E, Coskun B. Lowering the dose in head CT using adaptive statistical iterative reconstruction. AJNR Am J Neuroradiol 2011;32(9):1578–82. [9] Korn A, Fenchel M, Bender B, et al. Iterative reconstruction in head CT: image quality of routine and low-dose protocols in comparison with standard filtered back-projection. AJNR Am J Neuroradiol 2012;33(2):218–24. [10] Niu Y, Wang Z, Liu Y, Liu Z, Yao V. Radiation dose to the lens using different temporal bone CT scanning protocols. AJNR Am J Neuroradiol 2010;31(2):226–9. [11] Tumur O, Soon K, Brown F, Mykytowycz M. New scanning technique using adaptive statistical iterative reconstruction (ASIR) significantly reduced the radiation dose of cardiac CT. J Med Imaging Radiat Oncol 2013;57(3):292–6. [12] Gebhard C, Fuchs TA, Fiechter M, et al. Image quality of low-dose CCTA in obese patients: impact of high-definition computed tomography and adaptive statistical iterative reconstruction. Int J Cardiovasc Imaging 2013 [Epub ahead of print]. [13] Gebhard C, Fiechter M, Fuchs TA, et al. Coronary artery stents: influence of adaptive statistical iterative reconstruction on image quality using 64-HDCT. Eur Heart J Cardiovasc Imaging 2013 [Epub ahead of print]. [14] Fuchs TA, Fiechter M, Gebhard C, et al. CT coronary angiography: impact of adapted statistical iterative reconstruction (ASIR) on coronary stenosis and plaque composition analysis. Int J Cardiovasc Imaging 2013;29(3):719–24. [15] Gebhard C, Fiechter M, Fuchs TA, et al. Coronary artery calcium scoring: influence of adaptive statistical iterative reconstruction using 64-MDCT. Int J Cardiol 2012 [Epub ahead of print]. [16] Pontana F, Duhamel A, Pagniez J, et al. Chest computed tomography using iterative reconstruction vs filtered back projection (Part 2): image quality of low-dose CT examinations in 80 patients. Eur Radiol 2011;21(3):636–43. [17] Pontana F, Pagniez J, Flohr T, et al. Chest computed tomography using iterative reconstruction vs filtered back projection (Part 1): evaluation of image noise reduction in 32 patients. Eur Radiol 2011;21(3):627–35. [18] Singh S, Kalra MK, Gilman MD, et al. Adaptive statistical iterative reconstruction technique for radiation dose reduction in chest CT: a pilot study. Radiology 2011;259(2):565–73. [19] Shuman WP, Green DE, Busey JM, et al. Model-based iterative reconstruction versus adaptive statistical iterative reconstruction and filtered back projection in liver 64-MDCT: focal lesion detection, lesion conspicuity, and image noise. AJR Am J Roentgenol 2013;200(5):1071–6. [20] Singh S, Kalra MK, Hsieh J, et al. Abdominal CT: comparison of adaptive statistical iterative and filtered back projection reconstruction techniques. Radiology 2010;257(2):373–83. [21] The 2007 Recommendations of the International Commission on Radiological Protection. ICRP publication 103. Ann ICRP 2007;37(2–4):1–332. [22] Deak PD, Smal Y, Kalender WA. Multisection CT protocols: sex- and age-specific conversion factors used to determine effective dose from dose-length product. Radiology 2010;257(1):158–66. [23] Kalender W. Computed tomography: fundamentals, system technology, image quality, applications. Erlangen: Publicis Publishing; 2011. [24] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1):159–74.