- Email: [email protected]
Radiation Dose to the Breast by 64-slice CT
ARTICLE IN PRESS
Original Investigation
Radiation Dose to the Breast by 64-slice CT: Effects of Scanner Model and Study Protocol Zhihua Qi, PhD, Lisa C. Lemen, PhD, Michael Lamba, PhD, Hua-Hsuan Chen, PhD, Ranasinghage Samaratunga, PhD, Mary Mahoney, MD, R. Edward Hendrick, PhD Rationale and Objectives: This work aimed to study the effects of scanner model and study protocol on radiation dose received by breast tissues from 64-slice computed tomography (CT) studies. Materials and Methods: Four scanner models and three study protocols were used in scanning an anthropomorphic phantom with breast modules. Each protocol follows recommendations or guidelines from the American Association of Physicists in Medicine and the American College of Radiology. Twenty thermoluminescent dosimeters were placed inside the breast modules to measure breast tissue doses. Both the absolute and the normalized breast tissue doses were analyzed. Results: The mean glandular doses of a lung cancer screening CT, a chest/abdomen/pelvis CT, and a virtual colonoscopy CT are equivalent to less than 1, 5–7, and 1–3 two-view digital mammograms, respectively, for a standard-sized patient. The normalized breast dose differs significantly (P < 0.01) between lung cancer screening CT and chest/abdomen/pelvis CT; however, it shows less than ±10% variation among scanner models for the same protocol. In virtual colonoscopy CT, breast tissue dose decreases with the distance between local tissues to the edge of the x-ray field, although the decreasing trend varies for different scanner models and protocol settings. Conclusions: When breasts are entirely included in the primary x-ray field, breast dose by 64-slice CT is mainly protocol dependent, with the normalized breast dose about 15% lower for protocols with modulated mA than for those with constant mA; when breasts are only partially included in the primary beam field, breast dose by 64-slice CT is dependent on both the scanner model and the protocol settings. Key Words: Computed tomography; radiation dosimetry; breast dose. © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
I
n recent years, public concerns over computed tomography (CT) radiation dose and its associated risks have spurred efforts to understand, manage, and optimize patient dose from CT studies (1–4). Among these efforts, the recording and reporting of patient dose play an important role for patient management in clinical practice. Although CT dose index (CTDI), a metric directly available from scanners, has long been used in reporting scanner radiation output, its practical value is limited because of its inability to account for patient variation. Organ dose has been considered a more valuable and suitable metric to meet clinical needs; it gives physical Acad Radiol 2016; ■:■■–■■ From the Department of Radiology, University of Cincinnati, 234 Goodman Street PO Box 670761 Cincinnati, OH 45267-0761 (Z.Q., L.C.L., M.L., H.-H.C., R.S., M.M.); Department of Radiology, University of Colorado – Denver, School of Medicine, Aurora, CO 80045 (R.E.H.). Received January 14, 2016; revised March 4, 2016; accepted March 5, 2016. Address correspondence to: Z.Q. e-mail: [email protected] © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.acra.2016.03.022
characterization of patient-specific radiation dose and forms the basis for risk estimates. The determination of organ dose is a challenging task. The organ is affected not only by the scanner radiation output level, usually characterized by CTDI, but also by various other factors, such as patient size, study protocol, scanner model, and x-ray energy spectrum. A number of studies have been conducted with the aim of developing a robust yet simple method with acceptable accuracy for organ dose estimation (5–18). These studies employed either experimental methods using physical phantoms or numerical methods using validated Monte Carlo programs, although the majority of them were based on the latter because of the flexibility of simulation. This paper attempts to characterize CT-induced breast dose with a phantom-based experimental study. In recently published International Commission on Radiological Protection Publication 103 (19), the weighting factor for breast tissue in calculating patient risks was increased from 0.005 to 0.12, which puts an emphasis on breast dose in estimating radiationinduced patient risk. Although the dependence of breast dose 1
Academic Radiology, Vol ■, No ■, ■■ 2016
QI ET AL
MATERIALS AND METHODS Phantom and CT Scan Protocols
The Rando-Alderson anthropomorphic phantom (The Phantom Laboratory, Salem, NY, USA) with breast modules (Fig 1) was scanned on four 64-slice CT scanner models, including GE LightSpeed VCT, Siemens SOMATOM Sensation 64, Philips Brilliance 64, and Toshiba Aquilion 64. The breast modules have a typical anatomically relevant relaxed shape. Three types of CT studies were investigated in this study, including:
Figure 1. The Rando-Alderson phantom with breast modules attached to simulate a standard-sized female patient.
on patient size has been studied in the past (18), this paper focuses on the effects of scanner model and study protocol on breast dose, with the goal of providing guidance on how to effectively account for these two factors in the determination of breast dose from commonly performed CT studies.
(1) Lung cancer screening CT—this commonly used lowdose screening study represents those that typically use constant mA in scanning the chest region; (2) Chest/abdomen/pelvis CT—this most commonly used body CT study represents those that typically use tube current modulation technique in optimizing patient dose; and (3) Virtual colonoscopy CT—this study represents those in which breast tissues are in close vicinity of the scan coverage to receive nonnegligible x-ray scatter and may be partially included in the scan coverage to get direct exposure. The acquisition parameters of the protocols used in the study are summarized in Tables 1–3. They were primarily based on
TABLE 1. Acquisition Parameters for the Lung Screening CT Protocols Used on Different Systems
kVp mAs Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
GE LightSpeed VCT
Siemens SOMATOM Sensation 64
Philips Brilliance 64
Toshiba Aquilion 64
120 20 0.5 0.969 32 × 0.625 mm Large body 1.9
120 25 0.5 1.0 64 × 0.6 mm* Body 1.8
120 18 0.75 0.673 64 × 0.625 mm Body 1.7
120 15 0.5 0.828 64 × 0.5 mm Large 2.1
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
TABLE 2. Acquisition Parameters for the Chest/Abdomen/Pelvis CT Protocols Used on Different Systems GE LightSpeed VCT kVp Tube current modulation technique Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
120 Auto mA and Smart mA 0.4 0.984 64 × 0.625 mm Large body 20.0
Siemens SOMATOM Sensation 64 120 Care Dose 4D 0.5 0.9 64 × 0.6 mm* Body 20.0
Philips Brilliance 64
Toshiba Aquilion 64
120 D-DOM and Z-DOM 0.75 0.891 64 × 0.625 mm Body 20.0
120 Sure Exposure 0.5 0.828 64 × 0.5 mm Large 20.0
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
2
Academic Radiology, Vol ■, No ■, ■■ 2016
BREAST DOSE BY 64-SLICE CT
what is routinely used for clinical patients. For lung cancer screening CT and chest/abdomen/pelvis CT, the volume CTDI (CTDIvol) values were set to be equal to those in the CT protocols recommended by the American Association of Physicists in Medicine (20). For virtual colonoscopy CT, the CTDIvol values were set to be 10 mGy, 80% of the CTDIvol limit suggested by the American College of Radiology guidelines for dual-position CT virtual colonoscopy studies (21). Standardizing the radiation output of these protocols based on well-accepted references allows for an objective comparison of the characteristics of different scanners and the incurred
radiation dose to breast tissues. The scan lengths used in the study for the three protocols were indicated in Fig 3. Breast Dose Measurements by Thermoluminescent Dosimeter (TLD)
Absorbed dose to breast tissues was measured using LiFtype TLDs distributed throughout the breast modules. A total of 20 TLDs were used, with 11 of them placed in the right module and nine in the left module (Fig 2). In both lung cancer screening CT and chest/abdomen/pelvis CT, the breast
Figure 2. The diagram of the locations of the 20 thermoluminescent dosimeters (TLDs) (marked by circles) placed inside the breast modules, visualized in an anteriorposterior view.
Figure 3. Scan lengths used for three computed tomography (CT) protocols in the study.
TABLE 3. Acquisition Parameters for the Virtual Colonoscopy CT Protocols Used on Different Systems
kVp mAs (supine position) mAs (prone position) Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
GE LightSpeed VCT
Siemens SOMATOM Sensation 64
120 100 100 0.5 1.375 64 × 0.625 mm Large body 10.0
120 150 25 0.5 1.4 32 × 0.6 mm* Body 10.0
Philips Brilliance 64
Toshiba Aquilion 64
120 100 100 0.75 0.891 64 × 0.625 mm Body 10.0
120 75 75 0.5 1.48 64 × 0.5 mm Large 10.0
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
3
Academic Radiology, Vol ■, No ■, ■■ 2016
QI ET AL
modules are completely in the scan coverage; in virtual colonoscopy CT scans, only the most inferior row of TLDs (row 1 in Fig 2) was in the scan coverage. TLDs were read with a Harshaw 3500 single chip reader (Thermo Electron Corporation, Santa Fe, NM) annealed at 400°C for 1 hour and at 100°C for 2 hours. To convert the readings to meaningful physical doses, the TLDs were calibrated using a radiographic/fluoroscopic x-ray tube with additional filtration added to achieve a half-value layer of 9 mm of aluminum (120 kVp, 0.28 mm Cu added). Entrance dose of the calibration beam was measured with a National Institute of Standards and Technology traceable ion chamber. The calibration factor obtained in such a way converts the raw TLD readings into water-equivalent absorbed dose values. After the calibration, an additional conversion factor accounting for differences in attenuation characteristics between water and fibroglandular tissues was applied to obtain breast tissue– absorbed radiation dose. Data Analysis
Two dose metrics were used in data analysis. The first is the absolute breast tissue dose, in the unit of mGy. Its average value over the 20 TLD locations is considered as the mean glandular dose and compared to that from a typical twoview digital screening mammogram, which results in about 3.7 mGy mean glandular dose (22). The second is the normalized breast tissue dose, defined as the absolute breast tissue dose divided by the regional CTDIvol (14), as follows:
Normalized breast dose =
Absolute breast dose , Regional CTDIvol
which characterizes the resulting breast tissue dose per unit radiation output to the breasts, and allows comparison among protocols with different radiation output levels. The regional CTDIvol is defined as follows: CTDIvol ,regional = CTDIvol ,global ×
mean mA for image slicess containing breast tissues , mean mA for all image slices
where CTDIvol,global is the displayed CTDIvol from the scanner, representative of the dose index averaged over the entire scanner anatomy. For protocols using tube current modulation, CTDIvol,regional is more accurate than CTDIvol,global in characterizing the radiation output used on breast tissues. For protocols using constant mA, CTDIvol,regional becomes equal to CTDIvol,global. For lung cancer screening CT and chest/abdomen/pelvis CT, both the absolute and the normalized breast tissue doses were evaluated and compared across different scanner models and protocols. Because these two protocols use two different scan modes, constant mA vs modulated mA, the difference in the normalized breast tissue doses between the two protocols was assessed with a paired Student t test, with the null hypothesis that there is no difference between the two protocols in the normalized breast dose. A P value below 0.05 is considered to be statistically significant to reject the null hypothesis (Fig 3). 4
Figure 4. Breast doses on four scanner models for three computed tomography (CT) study protocols: lung cancer screening CT (a), chest/abdomen/pelvis CT (b), and virtual colonoscopy CT (c). The first two follow American Association of Physicists in Medicine (AAPM)-recommended protocol settings, and the third follows the American College of Radiology (ACR) guideline.
The virtual colonoscopy CT protocol differs from the other two because breast tissues are only partially included in the primary x-ray field. It is expected that breast tissue doses at different locations will exhibit a larger variation than those of lung cancer screening CT and chest/abdomen/pelvis CT protocols. In addition to the comparison of the normalized
Academic Radiology, Vol ■, No ■, ■■ 2016
BREAST DOSE BY 64-SLICE CT
Figure 5. The normalized breast tissue dose on four scanner models for the three studied protocols: lung cancer screening, chest/abdomen/pelvis, and virtual colonoscopy. Across the four 64-slice computed tomography (CT) scanner models, the normalized dose averaged over the entire breasts falls within ±5% of 1.05 and ±10% of 0.90 for the lung cancer screening protocol and the chest/abdomen/pelvis protocol, respectively. For the virtual colonoscopy protocol, however, both larger differences in the normalized dose among different scanner models and larger error bars are observed; this is primarily because of the fact that the breasts are only partially included in the primary x-ray field for such studies. For this reason, additional analysis of thermoluminescent dosimeter (TLD) data was performed for the virtual colonoscopy protocol, with the result shown in Figure 6.
breast tissue doses, the absolute breast tissue dose values were also analyzed as a function of the distances of the TLDs relative to the most inferior row of TLDs, which were just inside the scan coverage. RESULTS The absolute breast doses for the four scanner models and three study protocols are shown in Figure 4. The mean fibroglandular dose for a lung cancer screening CT study is below 3 mGy, less than that of a single two-view digital mammogram. The mean glandular dose for a chest/abdomen/pelvis CT is about that of 5–6 two-view digital mammograms. The mean glandular dose for a virtual colonoscopy CT shows larger variations across scanner models, ranging from the dose levels of 1–3 two-view digital mammograms. The normalized breast tissue doses for lung cancer screening CT studies and chest/abdomen/pelvis CT on four scanner models are shown in Figure 5. First, for each protocol, the normalized breast tissue dose averaged over the entire breasts shows a small range among different scanner models, ie, this value falls within ±5% of 1.05 and ±10% of 0.90, for lung cancer screening CT and chest/abdomen/pelvis CT, respectively. This finding indicates that, despite the scanner models’ differences in the characteristics of their tube current modulation techniques, their impacts on breast dose are similar to each other. Second, across different scanner models, it is consistently found that the normalized breast tissue dose of lung cancer screening CT is significantly higher (P < 0.01) than that of chest/abdomen/pelvis CT. This difference is about 15% when averaged over all four scanner models. Considering that
these two protocols represent those that use different mA settings, constant mA vs modulated mA, the observed difference indicates that protocols with modulated mA result in less breast dose per unit regional radiation output (CTDIvol,regional) than those with constant mA. This may be explained by the fact that Anterior-Posterior (AP) views result in more breast dose per unit mA than the other views, and tube current modulation tends to reduce mA in the AP views while increasing mA in the lateral views to achieve desired noise characteristics. For virtual colonoscopy CT, the normalized breast tissue doses for virtual colonoscopy CT are included in Figure 5. Both larger differences in the normalized dose among different scanner models and larger error bars are observed; this is within our expectation, owing to the fact that the breasts are only partially included in the primary x-ray field for such studies and their absorbed dose is from both primary and scattered radiations. In addition, the absolute breast tissue dose as a function of the TLD’s distance relative to the most inferior row of TLDs is plotted in Figure 6 for all scanner models. Out of the four scanner models, two show a linear decrease in dose with respect to the distance, and the other two scanners show an exponential decay. Overranging is believed to be the major contributing factor to the observed difference between the two groups. Overranging is the extension of scan length in helical CT beyond both ends of the imaged volume, and the two most important technical parameters that affect overranging are pitch and detector collimation. As is shown in Table 3, the two scanner models that show the exponential decay trend in Figure 6, ie, the Siemens SOMATOM Sensation 64 and the Philips Brilliance 64, use either a smaller pitch (0.891) or a smaller total detector collimation 5
QI ET AL
Figure 6. The breast doses measured by thermoluminescent dosimeters (TLDs) in the scans using the virtual colonoscopy protocol as a function of their z-distances with respect to the most inferior row of TLDs (also the ones included in the scan coverage). For two of the scanner models, the function can be approximated by a linear curve (correlation coefficient 0.9357). For the other two models, the function can be approximated by an exponential curve (correlation coefficient 0.9045).
(32 × 0.6 mm) compared to the other two that show the linear decreasing trend.
DISCUSSION For CT studies that include breasts entirely in their primary beam field, the normalized breast tissue dose shows significant differences between protocols with constant mA and those with modulated mA; for the same protocol, however, its variation across different 64-slice CT scanner models is small (no more than ±10%). This indicates that, for any CT protocol directly involving the breast, its resulting mean glandular dose on a standard-sized patient can be estimated with accuracy 6
Academic Radiology, Vol ■, No ■, ■■ 2016
and robustness on 64-slice CT, regardless of its scanner model, once its CTDIvol,regional and mA modulation strategy are known. For CT studies that include breasts partially in their primary beam field, like virtual colonoscopy CT, the resulting breast dose varies with both the scanner model, including supported detector collimation, and the protocol settings, including helical pitch. Ideally, breasts should be excluded from the primary beam field for protocols like colonoscopy; however, because of their close proximity to the anatomy of study and the overranging factor of 64-slice CT (23), it is very likely that some breast tissue will be included in the primary beam field even when the breasts are pulled away from the field during patient preparation. There are limitations in the presented work. First, the distribution of TLDs in the two breast modules (left vs right) was asymmetric, which affects the statistical analysis. It was caused by the physical limitation of the phantom design, ie, the two modules have unequal number of holes for TLD placement. Second, patient size, a major factor that affects breast dose, was not studied. The dependence of organ dose on patient size has been extensively studied using validated Monte Carlo simulation. A number of groups reported that CTDIvol normalized organ dose as a function of patient size can be approximated as an exponential curve (11,24). Their findings can be combined with our findings here to develop a complete methodology for accurate and robust determination of breast dose from CT studies. A breast shield was not used in either the lung screening or the chest/abdomen/pelvis CT examinations in this study. American Association of Physicists in Medicine recommendations were followed in making these protocols for the study, and the use of a bismuth breast shield for breast dose reduction is not recommended owing to concerns of compromised image quality. When a breast shield is used, a correction factor needs to be determined to account for the attenuation by the shield for breast dose estimation. CONCLUSION The resulting mean glandular doses of a lung cancer screening CT, a chest/abdomen/pelvis CT, and a virtual colonoscopy CT are equivalent to less than 1, 5–7, and 1–3 two-view digital mammograms, respectively, for a standard-sized patient. When breasts are entirely included in the primary x-ray field, breast dose by 64-slice CT is mainly protocol dependent, with the normalized breast dose about 15% lower for protocols with modulated mA than for those with constant mA; when breasts are only partially included in the primary beam field, breast dose by 64-slice CT is dependent on both the scanner model and the protocol settings. ACKNOWLEDGMENT We thank GE, Philips, Siemens, and Toshiba personnel for their assistance and interest in an independent assessment. We gratefully acknowledge the following institutions for
Academic Radiology, Vol ■, No ■, ■■ 2016
providing clinical expertise and allowing us to use their scanners for this project: University of Cincinnati Physicians Medical Arts Building, Cincinnati Children’s Hospital Medical Center, Jewish Hospital, and ProScan Imaging Midtown. We also thank Jane Abbottsmith for her technical assistance. REFERENCES 1. Goske MJ, Applegate KE, Boylan J, et al. Image gently: a national education and communication campaign in radiology using the science of social marketing. J Am Coll Radiol 2008; 5:1200–1205. 2. Brink JA, Amis ES, Jr. Image wisely: a campaign to increase awareness about adult radiation protection. Radiology 2010; 257:601–602. 3. Boone JM, Hendee WR, McNitt-Gray MF, et al. Radiation exposure from CT scans: how to close our knowledge gaps, monitor and safeguard exposure. Radiology 2012; 265:544–554. 4. McCollough CH, Chen GH, Kalender W, et al. Achieving routine submillisievert CT scanning: report from the summit on management of radiation dose in CT. Radiology 2012; 264:567–580. 5. Fujii K, Aoyama T, Yamauchi-Kawaura C, et al. Radiation dose evaluation in 64-slice CT examinations with adult and paediatric anthropomorphic phantoms. Br J Radiol 2009; 82:1010–1018. 6. DeMarco JJ, Cagnon CH, Cody DD, et al. Estimating radiation doses from multidetector CT using Monte Carlo simulations: effects of different size voxelized patient models on magnitudes of organ and effective dose. Phys Med Biol 2007; 52:2583–2597. 7. Angel E, Yaghmai N, Jude CM, et al. Dose to radiosensitive organs during routine chest CT: effects of tube current modulation. AJR Am J Roentgenol 2009; 193:1340–1345. 8. Huda W, Sterzik A, Tipnis S, et al. Organ doses to adult patients for chest CT. Med Phys 2010; 37:842–847. 9. Turner AC, Zhang D, Khatonabadi M, et al. The feasibility of a scannerindependent technique to estimate organ dose from MDCT scans: using CTDIvol to account for differences between scanners. Med Phys 2010; 37:1816–1825. 10. Li X, Samei E, Segars WP, et al. Patient-specific radiation dose and cancer risk estimation in CT: part II. Application to patients. Med Phys 2011; 38:408–419.
BREAST DOSE BY 64-SLICE CT
11. Turner AC, Zhang D, Khatonabadi M, et al. The feasibility of patient sizecorrected, scanner-independent organ dose estimates for abdominal CT exams. Med Phys 2011; 38:820–829. 12. Lee C, Kim KP, Long DJ, et al. Organ doses for reference pediatric and adolescent patients undergoing computed tomography estimated by Monte Carlo simulation. Med Phys 2012; 39:2129–2146. 13. Li X, Samei E, Williams CH, et al. Effects of protocol and obesity on dose conversion factors in adult body CT. Med Phys 2012; 39:6550– 6571. 14. Khatonabadi M, Kim HJ, Lu P, et al. The feasibility of a regional CTDIvol to estimate organ dose from tube current modulated CT exams. Med Phys 2013; 40:051903. 15. Tian X, Li X, Segars WP, et al. Dose coefficients in pediatric and adult abdominopelvic CT based on 100 patient models. Phys Med Biol 2013; 58:8755–8768. 16. Li X, Segars WP, Samei E. The impact on CT dose of the variability in tube current modulation technology: a theoretical investigation. Phys Med Biol 2014; 59:4525–4548. 17. Sahbaee P, Segars WP, Samei E. Patient-based estimation of organ dose for a population of 58 adult patients across 13 protocol categories. Med Phys 2014; 41:072104. 18. Angel E, Yaghmai N, Jude CM, et al. Monte Carlo simulations to assess the effects of tube current modulation on breast dose for multidetector CT. Phys Med Biol 2009; 54:497–511. 19. The 2007 Recommendations of the International Commission on Radiological Protection. ICRP Publication 103. Ann ICRP 2007; 37: 2–4. 20. AAPM. CT scan protocols. 2015. Available at: http://www.aapm.org/pubs/ CTProtocols/default.asp. 21. ACR-SAR-SCBT-MR practice parameter for the performance of computed tomography colonoscopy in adults. 2014. Available at: http://www.acr.org. 22. Hendrick RE. Radiation doses and cancer risks from breast imaging studies. Radiology 2010; 257:246–253. 23. Schilham A, van der Molen AJ, Prokop M, et al. Overranging at multisection CT: an underestimated source of excess radiation exposure. Radiographics 2010; 30:1057–1067. 24. Tian X, Li X, Segars WP, et al. Pediatric chest and abdominopelvic CT: organ dose estimation based on 42 patient models. Radiology 2014; 270:535–547.
7
Original Investigation
Radiation Dose to the Breast by 64-slice CT: Effects of Scanner Model and Study Protocol Zhihua Qi, PhD, Lisa C. Lemen, PhD, Michael Lamba, PhD, Hua-Hsuan Chen, PhD, Ranasinghage Samaratunga, PhD, Mary Mahoney, MD, R. Edward Hendrick, PhD Rationale and Objectives: This work aimed to study the effects of scanner model and study protocol on radiation dose received by breast tissues from 64-slice computed tomography (CT) studies. Materials and Methods: Four scanner models and three study protocols were used in scanning an anthropomorphic phantom with breast modules. Each protocol follows recommendations or guidelines from the American Association of Physicists in Medicine and the American College of Radiology. Twenty thermoluminescent dosimeters were placed inside the breast modules to measure breast tissue doses. Both the absolute and the normalized breast tissue doses were analyzed. Results: The mean glandular doses of a lung cancer screening CT, a chest/abdomen/pelvis CT, and a virtual colonoscopy CT are equivalent to less than 1, 5–7, and 1–3 two-view digital mammograms, respectively, for a standard-sized patient. The normalized breast dose differs significantly (P < 0.01) between lung cancer screening CT and chest/abdomen/pelvis CT; however, it shows less than ±10% variation among scanner models for the same protocol. In virtual colonoscopy CT, breast tissue dose decreases with the distance between local tissues to the edge of the x-ray field, although the decreasing trend varies for different scanner models and protocol settings. Conclusions: When breasts are entirely included in the primary x-ray field, breast dose by 64-slice CT is mainly protocol dependent, with the normalized breast dose about 15% lower for protocols with modulated mA than for those with constant mA; when breasts are only partially included in the primary beam field, breast dose by 64-slice CT is dependent on both the scanner model and the protocol settings. Key Words: Computed tomography; radiation dosimetry; breast dose. © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
I
n recent years, public concerns over computed tomography (CT) radiation dose and its associated risks have spurred efforts to understand, manage, and optimize patient dose from CT studies (1–4). Among these efforts, the recording and reporting of patient dose play an important role for patient management in clinical practice. Although CT dose index (CTDI), a metric directly available from scanners, has long been used in reporting scanner radiation output, its practical value is limited because of its inability to account for patient variation. Organ dose has been considered a more valuable and suitable metric to meet clinical needs; it gives physical Acad Radiol 2016; ■:■■–■■ From the Department of Radiology, University of Cincinnati, 234 Goodman Street PO Box 670761 Cincinnati, OH 45267-0761 (Z.Q., L.C.L., M.L., H.-H.C., R.S., M.M.); Department of Radiology, University of Colorado – Denver, School of Medicine, Aurora, CO 80045 (R.E.H.). Received January 14, 2016; revised March 4, 2016; accepted March 5, 2016. Address correspondence to: Z.Q. e-mail: [email protected] © 2016 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.acra.2016.03.022
characterization of patient-specific radiation dose and forms the basis for risk estimates. The determination of organ dose is a challenging task. The organ is affected not only by the scanner radiation output level, usually characterized by CTDI, but also by various other factors, such as patient size, study protocol, scanner model, and x-ray energy spectrum. A number of studies have been conducted with the aim of developing a robust yet simple method with acceptable accuracy for organ dose estimation (5–18). These studies employed either experimental methods using physical phantoms or numerical methods using validated Monte Carlo programs, although the majority of them were based on the latter because of the flexibility of simulation. This paper attempts to characterize CT-induced breast dose with a phantom-based experimental study. In recently published International Commission on Radiological Protection Publication 103 (19), the weighting factor for breast tissue in calculating patient risks was increased from 0.005 to 0.12, which puts an emphasis on breast dose in estimating radiationinduced patient risk. Although the dependence of breast dose 1
Academic Radiology, Vol ■, No ■, ■■ 2016
QI ET AL
MATERIALS AND METHODS Phantom and CT Scan Protocols
The Rando-Alderson anthropomorphic phantom (The Phantom Laboratory, Salem, NY, USA) with breast modules (Fig 1) was scanned on four 64-slice CT scanner models, including GE LightSpeed VCT, Siemens SOMATOM Sensation 64, Philips Brilliance 64, and Toshiba Aquilion 64. The breast modules have a typical anatomically relevant relaxed shape. Three types of CT studies were investigated in this study, including:
Figure 1. The Rando-Alderson phantom with breast modules attached to simulate a standard-sized female patient.
on patient size has been studied in the past (18), this paper focuses on the effects of scanner model and study protocol on breast dose, with the goal of providing guidance on how to effectively account for these two factors in the determination of breast dose from commonly performed CT studies.
(1) Lung cancer screening CT—this commonly used lowdose screening study represents those that typically use constant mA in scanning the chest region; (2) Chest/abdomen/pelvis CT—this most commonly used body CT study represents those that typically use tube current modulation technique in optimizing patient dose; and (3) Virtual colonoscopy CT—this study represents those in which breast tissues are in close vicinity of the scan coverage to receive nonnegligible x-ray scatter and may be partially included in the scan coverage to get direct exposure. The acquisition parameters of the protocols used in the study are summarized in Tables 1–3. They were primarily based on
TABLE 1. Acquisition Parameters for the Lung Screening CT Protocols Used on Different Systems
kVp mAs Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
GE LightSpeed VCT
Siemens SOMATOM Sensation 64
Philips Brilliance 64
Toshiba Aquilion 64
120 20 0.5 0.969 32 × 0.625 mm Large body 1.9
120 25 0.5 1.0 64 × 0.6 mm* Body 1.8
120 18 0.75 0.673 64 × 0.625 mm Body 1.7
120 15 0.5 0.828 64 × 0.5 mm Large 2.1
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
TABLE 2. Acquisition Parameters for the Chest/Abdomen/Pelvis CT Protocols Used on Different Systems GE LightSpeed VCT kVp Tube current modulation technique Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
120 Auto mA and Smart mA 0.4 0.984 64 × 0.625 mm Large body 20.0
Siemens SOMATOM Sensation 64 120 Care Dose 4D 0.5 0.9 64 × 0.6 mm* Body 20.0
Philips Brilliance 64
Toshiba Aquilion 64
120 D-DOM and Z-DOM 0.75 0.891 64 × 0.625 mm Body 20.0
120 Sure Exposure 0.5 0.828 64 × 0.5 mm Large 20.0
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
2
Academic Radiology, Vol ■, No ■, ■■ 2016
BREAST DOSE BY 64-SLICE CT
what is routinely used for clinical patients. For lung cancer screening CT and chest/abdomen/pelvis CT, the volume CTDI (CTDIvol) values were set to be equal to those in the CT protocols recommended by the American Association of Physicists in Medicine (20). For virtual colonoscopy CT, the CTDIvol values were set to be 10 mGy, 80% of the CTDIvol limit suggested by the American College of Radiology guidelines for dual-position CT virtual colonoscopy studies (21). Standardizing the radiation output of these protocols based on well-accepted references allows for an objective comparison of the characteristics of different scanners and the incurred
radiation dose to breast tissues. The scan lengths used in the study for the three protocols were indicated in Fig 3. Breast Dose Measurements by Thermoluminescent Dosimeter (TLD)
Absorbed dose to breast tissues was measured using LiFtype TLDs distributed throughout the breast modules. A total of 20 TLDs were used, with 11 of them placed in the right module and nine in the left module (Fig 2). In both lung cancer screening CT and chest/abdomen/pelvis CT, the breast
Figure 2. The diagram of the locations of the 20 thermoluminescent dosimeters (TLDs) (marked by circles) placed inside the breast modules, visualized in an anteriorposterior view.
Figure 3. Scan lengths used for three computed tomography (CT) protocols in the study.
TABLE 3. Acquisition Parameters for the Virtual Colonoscopy CT Protocols Used on Different Systems
kVp mAs (supine position) mAs (prone position) Rotation time (s) Pitch Detector configuration Bow-tie filter CTDIvol (mGy)
GE LightSpeed VCT
Siemens SOMATOM Sensation 64
120 100 100 0.5 1.375 64 × 0.625 mm Large body 10.0
120 150 25 0.5 1.4 32 × 0.6 mm* Body 10.0
Philips Brilliance 64
Toshiba Aquilion 64
120 100 100 0.75 0.891 64 × 0.625 mm Body 10.0
120 75 75 0.5 1.48 64 × 0.5 mm Large 10.0
CT, computed tomography; CTDIvol, volume CT dose index. * The physical beam width is 32 × 0.6 mm. The flying focal spot technology is used to produce 64 slices per rotation.
3
Academic Radiology, Vol ■, No ■, ■■ 2016
QI ET AL
modules are completely in the scan coverage; in virtual colonoscopy CT scans, only the most inferior row of TLDs (row 1 in Fig 2) was in the scan coverage. TLDs were read with a Harshaw 3500 single chip reader (Thermo Electron Corporation, Santa Fe, NM) annealed at 400°C for 1 hour and at 100°C for 2 hours. To convert the readings to meaningful physical doses, the TLDs were calibrated using a radiographic/fluoroscopic x-ray tube with additional filtration added to achieve a half-value layer of 9 mm of aluminum (120 kVp, 0.28 mm Cu added). Entrance dose of the calibration beam was measured with a National Institute of Standards and Technology traceable ion chamber. The calibration factor obtained in such a way converts the raw TLD readings into water-equivalent absorbed dose values. After the calibration, an additional conversion factor accounting for differences in attenuation characteristics between water and fibroglandular tissues was applied to obtain breast tissue– absorbed radiation dose. Data Analysis
Two dose metrics were used in data analysis. The first is the absolute breast tissue dose, in the unit of mGy. Its average value over the 20 TLD locations is considered as the mean glandular dose and compared to that from a typical twoview digital screening mammogram, which results in about 3.7 mGy mean glandular dose (22). The second is the normalized breast tissue dose, defined as the absolute breast tissue dose divided by the regional CTDIvol (14), as follows:
Normalized breast dose =
Absolute breast dose , Regional CTDIvol
which characterizes the resulting breast tissue dose per unit radiation output to the breasts, and allows comparison among protocols with different radiation output levels. The regional CTDIvol is defined as follows: CTDIvol ,regional = CTDIvol ,global ×
mean mA for image slicess containing breast tissues , mean mA for all image slices
where CTDIvol,global is the displayed CTDIvol from the scanner, representative of the dose index averaged over the entire scanner anatomy. For protocols using tube current modulation, CTDIvol,regional is more accurate than CTDIvol,global in characterizing the radiation output used on breast tissues. For protocols using constant mA, CTDIvol,regional becomes equal to CTDIvol,global. For lung cancer screening CT and chest/abdomen/pelvis CT, both the absolute and the normalized breast tissue doses were evaluated and compared across different scanner models and protocols. Because these two protocols use two different scan modes, constant mA vs modulated mA, the difference in the normalized breast tissue doses between the two protocols was assessed with a paired Student t test, with the null hypothesis that there is no difference between the two protocols in the normalized breast dose. A P value below 0.05 is considered to be statistically significant to reject the null hypothesis (Fig 3). 4
Figure 4. Breast doses on four scanner models for three computed tomography (CT) study protocols: lung cancer screening CT (a), chest/abdomen/pelvis CT (b), and virtual colonoscopy CT (c). The first two follow American Association of Physicists in Medicine (AAPM)-recommended protocol settings, and the third follows the American College of Radiology (ACR) guideline.
The virtual colonoscopy CT protocol differs from the other two because breast tissues are only partially included in the primary x-ray field. It is expected that breast tissue doses at different locations will exhibit a larger variation than those of lung cancer screening CT and chest/abdomen/pelvis CT protocols. In addition to the comparison of the normalized
Academic Radiology, Vol ■, No ■, ■■ 2016
BREAST DOSE BY 64-SLICE CT
Figure 5. The normalized breast tissue dose on four scanner models for the three studied protocols: lung cancer screening, chest/abdomen/pelvis, and virtual colonoscopy. Across the four 64-slice computed tomography (CT) scanner models, the normalized dose averaged over the entire breasts falls within ±5% of 1.05 and ±10% of 0.90 for the lung cancer screening protocol and the chest/abdomen/pelvis protocol, respectively. For the virtual colonoscopy protocol, however, both larger differences in the normalized dose among different scanner models and larger error bars are observed; this is primarily because of the fact that the breasts are only partially included in the primary x-ray field for such studies. For this reason, additional analysis of thermoluminescent dosimeter (TLD) data was performed for the virtual colonoscopy protocol, with the result shown in Figure 6.
breast tissue doses, the absolute breast tissue dose values were also analyzed as a function of the distances of the TLDs relative to the most inferior row of TLDs, which were just inside the scan coverage. RESULTS The absolute breast doses for the four scanner models and three study protocols are shown in Figure 4. The mean fibroglandular dose for a lung cancer screening CT study is below 3 mGy, less than that of a single two-view digital mammogram. The mean glandular dose for a chest/abdomen/pelvis CT is about that of 5–6 two-view digital mammograms. The mean glandular dose for a virtual colonoscopy CT shows larger variations across scanner models, ranging from the dose levels of 1–3 two-view digital mammograms. The normalized breast tissue doses for lung cancer screening CT studies and chest/abdomen/pelvis CT on four scanner models are shown in Figure 5. First, for each protocol, the normalized breast tissue dose averaged over the entire breasts shows a small range among different scanner models, ie, this value falls within ±5% of 1.05 and ±10% of 0.90, for lung cancer screening CT and chest/abdomen/pelvis CT, respectively. This finding indicates that, despite the scanner models’ differences in the characteristics of their tube current modulation techniques, their impacts on breast dose are similar to each other. Second, across different scanner models, it is consistently found that the normalized breast tissue dose of lung cancer screening CT is significantly higher (P < 0.01) than that of chest/abdomen/pelvis CT. This difference is about 15% when averaged over all four scanner models. Considering that
these two protocols represent those that use different mA settings, constant mA vs modulated mA, the observed difference indicates that protocols with modulated mA result in less breast dose per unit regional radiation output (CTDIvol,regional) than those with constant mA. This may be explained by the fact that Anterior-Posterior (AP) views result in more breast dose per unit mA than the other views, and tube current modulation tends to reduce mA in the AP views while increasing mA in the lateral views to achieve desired noise characteristics. For virtual colonoscopy CT, the normalized breast tissue doses for virtual colonoscopy CT are included in Figure 5. Both larger differences in the normalized dose among different scanner models and larger error bars are observed; this is within our expectation, owing to the fact that the breasts are only partially included in the primary x-ray field for such studies and their absorbed dose is from both primary and scattered radiations. In addition, the absolute breast tissue dose as a function of the TLD’s distance relative to the most inferior row of TLDs is plotted in Figure 6 for all scanner models. Out of the four scanner models, two show a linear decrease in dose with respect to the distance, and the other two scanners show an exponential decay. Overranging is believed to be the major contributing factor to the observed difference between the two groups. Overranging is the extension of scan length in helical CT beyond both ends of the imaged volume, and the two most important technical parameters that affect overranging are pitch and detector collimation. As is shown in Table 3, the two scanner models that show the exponential decay trend in Figure 6, ie, the Siemens SOMATOM Sensation 64 and the Philips Brilliance 64, use either a smaller pitch (0.891) or a smaller total detector collimation 5
QI ET AL
Figure 6. The breast doses measured by thermoluminescent dosimeters (TLDs) in the scans using the virtual colonoscopy protocol as a function of their z-distances with respect to the most inferior row of TLDs (also the ones included in the scan coverage). For two of the scanner models, the function can be approximated by a linear curve (correlation coefficient 0.9357). For the other two models, the function can be approximated by an exponential curve (correlation coefficient 0.9045).
(32 × 0.6 mm) compared to the other two that show the linear decreasing trend.
DISCUSSION For CT studies that include breasts entirely in their primary beam field, the normalized breast tissue dose shows significant differences between protocols with constant mA and those with modulated mA; for the same protocol, however, its variation across different 64-slice CT scanner models is small (no more than ±10%). This indicates that, for any CT protocol directly involving the breast, its resulting mean glandular dose on a standard-sized patient can be estimated with accuracy 6
Academic Radiology, Vol ■, No ■, ■■ 2016
and robustness on 64-slice CT, regardless of its scanner model, once its CTDIvol,regional and mA modulation strategy are known. For CT studies that include breasts partially in their primary beam field, like virtual colonoscopy CT, the resulting breast dose varies with both the scanner model, including supported detector collimation, and the protocol settings, including helical pitch. Ideally, breasts should be excluded from the primary beam field for protocols like colonoscopy; however, because of their close proximity to the anatomy of study and the overranging factor of 64-slice CT (23), it is very likely that some breast tissue will be included in the primary beam field even when the breasts are pulled away from the field during patient preparation. There are limitations in the presented work. First, the distribution of TLDs in the two breast modules (left vs right) was asymmetric, which affects the statistical analysis. It was caused by the physical limitation of the phantom design, ie, the two modules have unequal number of holes for TLD placement. Second, patient size, a major factor that affects breast dose, was not studied. The dependence of organ dose on patient size has been extensively studied using validated Monte Carlo simulation. A number of groups reported that CTDIvol normalized organ dose as a function of patient size can be approximated as an exponential curve (11,24). Their findings can be combined with our findings here to develop a complete methodology for accurate and robust determination of breast dose from CT studies. A breast shield was not used in either the lung screening or the chest/abdomen/pelvis CT examinations in this study. American Association of Physicists in Medicine recommendations were followed in making these protocols for the study, and the use of a bismuth breast shield for breast dose reduction is not recommended owing to concerns of compromised image quality. When a breast shield is used, a correction factor needs to be determined to account for the attenuation by the shield for breast dose estimation. CONCLUSION The resulting mean glandular doses of a lung cancer screening CT, a chest/abdomen/pelvis CT, and a virtual colonoscopy CT are equivalent to less than 1, 5–7, and 1–3 two-view digital mammograms, respectively, for a standard-sized patient. When breasts are entirely included in the primary x-ray field, breast dose by 64-slice CT is mainly protocol dependent, with the normalized breast dose about 15% lower for protocols with modulated mA than for those with constant mA; when breasts are only partially included in the primary beam field, breast dose by 64-slice CT is dependent on both the scanner model and the protocol settings. ACKNOWLEDGMENT We thank GE, Philips, Siemens, and Toshiba personnel for their assistance and interest in an independent assessment. We gratefully acknowledge the following institutions for
Academic Radiology, Vol ■, No ■, ■■ 2016
providing clinical expertise and allowing us to use their scanners for this project: University of Cincinnati Physicians Medical Arts Building, Cincinnati Children’s Hospital Medical Center, Jewish Hospital, and ProScan Imaging Midtown. We also thank Jane Abbottsmith for her technical assistance. REFERENCES 1. Goske MJ, Applegate KE, Boylan J, et al. Image gently: a national education and communication campaign in radiology using the science of social marketing. J Am Coll Radiol 2008; 5:1200–1205. 2. Brink JA, Amis ES, Jr. Image wisely: a campaign to increase awareness about adult radiation protection. Radiology 2010; 257:601–602. 3. Boone JM, Hendee WR, McNitt-Gray MF, et al. Radiation exposure from CT scans: how to close our knowledge gaps, monitor and safeguard exposure. Radiology 2012; 265:544–554. 4. McCollough CH, Chen GH, Kalender W, et al. Achieving routine submillisievert CT scanning: report from the summit on management of radiation dose in CT. Radiology 2012; 264:567–580. 5. Fujii K, Aoyama T, Yamauchi-Kawaura C, et al. Radiation dose evaluation in 64-slice CT examinations with adult and paediatric anthropomorphic phantoms. Br J Radiol 2009; 82:1010–1018. 6. DeMarco JJ, Cagnon CH, Cody DD, et al. Estimating radiation doses from multidetector CT using Monte Carlo simulations: effects of different size voxelized patient models on magnitudes of organ and effective dose. Phys Med Biol 2007; 52:2583–2597. 7. Angel E, Yaghmai N, Jude CM, et al. Dose to radiosensitive organs during routine chest CT: effects of tube current modulation. AJR Am J Roentgenol 2009; 193:1340–1345. 8. Huda W, Sterzik A, Tipnis S, et al. Organ doses to adult patients for chest CT. Med Phys 2010; 37:842–847. 9. Turner AC, Zhang D, Khatonabadi M, et al. The feasibility of a scannerindependent technique to estimate organ dose from MDCT scans: using CTDIvol to account for differences between scanners. Med Phys 2010; 37:1816–1825. 10. Li X, Samei E, Segars WP, et al. Patient-specific radiation dose and cancer risk estimation in CT: part II. Application to patients. Med Phys 2011; 38:408–419.
BREAST DOSE BY 64-SLICE CT
11. Turner AC, Zhang D, Khatonabadi M, et al. The feasibility of patient sizecorrected, scanner-independent organ dose estimates for abdominal CT exams. Med Phys 2011; 38:820–829. 12. Lee C, Kim KP, Long DJ, et al. Organ doses for reference pediatric and adolescent patients undergoing computed tomography estimated by Monte Carlo simulation. Med Phys 2012; 39:2129–2146. 13. Li X, Samei E, Williams CH, et al. Effects of protocol and obesity on dose conversion factors in adult body CT. Med Phys 2012; 39:6550– 6571. 14. Khatonabadi M, Kim HJ, Lu P, et al. The feasibility of a regional CTDIvol to estimate organ dose from tube current modulated CT exams. Med Phys 2013; 40:051903. 15. Tian X, Li X, Segars WP, et al. Dose coefficients in pediatric and adult abdominopelvic CT based on 100 patient models. Phys Med Biol 2013; 58:8755–8768. 16. Li X, Segars WP, Samei E. The impact on CT dose of the variability in tube current modulation technology: a theoretical investigation. Phys Med Biol 2014; 59:4525–4548. 17. Sahbaee P, Segars WP, Samei E. Patient-based estimation of organ dose for a population of 58 adult patients across 13 protocol categories. Med Phys 2014; 41:072104. 18. Angel E, Yaghmai N, Jude CM, et al. Monte Carlo simulations to assess the effects of tube current modulation on breast dose for multidetector CT. Phys Med Biol 2009; 54:497–511. 19. The 2007 Recommendations of the International Commission on Radiological Protection. ICRP Publication 103. Ann ICRP 2007; 37: 2–4. 20. AAPM. CT scan protocols. 2015. Available at: http://www.aapm.org/pubs/ CTProtocols/default.asp. 21. ACR-SAR-SCBT-MR practice parameter for the performance of computed tomography colonoscopy in adults. 2014. Available at: http://www.acr.org. 22. Hendrick RE. Radiation doses and cancer risks from breast imaging studies. Radiology 2010; 257:246–253. 23. Schilham A, van der Molen AJ, Prokop M, et al. Overranging at multisection CT: an underestimated source of excess radiation exposure. Radiographics 2010; 30:1057–1067. 24. Tian X, Li X, Segars WP, et al. Pediatric chest and abdominopelvic CT: organ dose estimation based on 42 patient models. Radiology 2014; 270:535–547.
7