Pediatr Radiol (2014) 44 (Suppl 3):S460–S467 DOI 10.1007/s00247-014-3117-7

IMAGE GENTLY ALARA CT SUMMIT: HOW TO USE NEW CT TECHNOLOGIES FOR CHILDREN

Determining organ dose: the holy grail Ehsan Samei & Xiaoyu Tian & W. Paul Segars

Received: 10 March 2014 / Revised: 29 April 2014 / Accepted: 8 July 2014 # Springer-Verlag Berlin Heidelberg 2014

Abstract Among the various metrics to quantify CT radiation dose, organ dose is generally regarded as one of the best to reflect patient radiation burden. Organ dose is dependent on two main factors, namely patient anatomy and irradiation field. An accurate estimation of organ dose requires detailed modeling of both factors. The modeling of patient anatomy needs to reflect the anatomical diversity and complexity across the population so that the attributes of a given clinical patient can be properly accounted for. The modeling of the irradiation field needs to accurately reflect the CT system condition, especially the tube current modulation (TCM) technique. We present an atlas-based method to model patient anatomy via a library of computational phantoms with representative ages, sizes and genders. A clinical patient is matched with a corresponding computational phantom to obtain a representation of patient anatomy. The irradiation field of the CT system is modeled using a validated Monte Carlo simulation program. The tube current modulation profiles are simulated using a manufacturer-generalizable ray-tracing algorithm. Combining the patient model, Monte Carlo results, and TCM profile, organ doses are obtained by multiplying organ dose values from a fixed mA scan (normalized to CTDIvol-normalized, denoted as horgan) and an adjustment factor that reflects the specific irradiation of each organ. The accuracy of the proposed method was quantified by simulating clinical abdominopelvic examinations of 58 patients. The predicted organ doses showed good agreement with simulated organ dose across all organs and modulation schemes. For an average CTDIvol of a CT exam of 10 mGy, the absolute median error across all organs was 0.64 mGy E. Samei (*) : X. Tian : W. P. Segars Carl E. Ravin Advanced Imaging Laboratories, Departments of Radiology, Biomedical Engineering, Physics, and Electrical Engineering, Duke University, 2424 Erwin Road, Suite 302, Durham, NC 27705, USA e-mail: [email protected]

(−0.21 and 0.97 for 25th and 75th percentiles, respectively). The percentage differences were within 15%. The study demonstrates that it is feasible to estimate organ doses in clinical CT examinations for protocols without and with tube current modulation. The methodology can be used for both prospective and retrospective estimation of organ dose. Keywords Computed tomography . Monte Carlo . Organ dose . Patient–specific . Pediatric

Introduction Computed tomography (CT) has become an indispensable diagnostic tool employed with great frequency [1]. Although this use is justified considering the great medical benefit from the use of CT, radiation dose at the population level has become a subject of public attention and concern [2]. Although the exact magnitude of risk has been debated, rightsizing the magnitude of radiation dose used for a clinical examination of any given patient has become a major goal for the CT community [3, 4]. Right-sizing implies a proper metric for radiation dose. To be effective, such a metric needs to accurately account for the radiation output of the scanner and the geometrical properties of a specific patient. Various metrics of radiation burden have been proposed over the years and used by the CT community. Table 1 provides a summary of currently available metrics for CT radiation exposure. As shown, each metric provides a different representation of the factors contributing to CT radiation burden. Each also has limitations. CTDIvol reflects the radiation output of a CT system in units of dose to a standard-size object. Although CTDIvol is effective in characterizing the system output for CT protocols with extended axial coverage, it fails to represent all protocols and to fully account for individual patient attributes [5]. The size-specific dose

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

S461

Table 1 Summary of currently available metrics for CT radiation exposure Metric

Representation

Factors accounted

Factors ignored or limitations

CTDIvol

Radiation output of a CT system

SSDE

Average radiation dose received by the patient assuming homogeneous tissue composition

Scanner model Scanning parameter Scanner model Scanning parameter Patient size

Organ dose

Dose to individual organs; could be estimated by Monte Carlo simulation or experimental measurement

Patient characteristics (e.g., patient size, organ location) Protocols with short axial coverage Patient anatomy (i.e. body shape, organ location and orientation, tissue composition) Organ radiation sensitivity Age and gender effect on radiation risk Protocols other than chest, abdomen, or pelvis Organ radiation sensitivity Age and gender effect on radiation risk Reflects the radiation burden in terms of multiple numbers for a single scan

Effective dose

Weighted sum of organ/tissue equivalent dose for radiation-sensitive organs with weighting factors defined by ICRP Weighted sum of organ/tissue equivalent dose for radiation-sensitive organs, with weighting factors based on lifetime cancer risk per unit organ dose

Risk index (RI)

Scanner model Scanning parameter Patient anatomy (i.e. body shape, organ location and orientation) Organ dose for whole body Organ radiation sensitivity Patient anatomy variation Organ dose for whole body Organ radiation sensitivity Age and gender effect on radiation risk

Patient anatomy variation (defined for an ICRP reference phantom) Age and gender effect on radiation risk Requires knowledge of organ dose RI is only an index, not an actual physical quantity. Can only serve as an index of risk overlooking uncertainties in the organ risk coefficients as well as other components of risk (e.g., genetic disposition)

ICRP International Commission on Radiological Protection, SSDE size-specific dose estimate, CTDI CT dose index

estimate (SSDE) incorporates the dependency of dose on a patient’s overall size but not the patient anatomy (i.e. body shape, organ location and orientation, and tissue composition), and SSDE shares some of the same limitations in protocol representation as CTDI [6]. A single size-specific dose estimate ascribed to an examination is associated with different doses to different organs. In contrast to CTDIvol and SSDE, organ dose addresses the effects of scanner output as well as patient characteristics. In doing so, it offers the most detailed reflection of radiation burden to the patient, compartmentalizing it into physically meaningful quantities. This level of granularity can be invaluable in clinical optimization. However, quantifying organ dose in a way that would truly reflect the dose to the organs of the actual patient, and not of an idealized object, has proved challenging. That is the reason for the continued use of alternatives, with organ dose determination considered the holy grail of patient dosimetry. It should be acknowledged that although organ dose provides the most meaningful reflection of radiation burden, it falls short of meeting the desire to provide a single value for the patient’s whole-body dose. Although a scalar metrology might seem over-simplistic, it is the most clinically useful for risk communication and for optimization purposes. Effective dose meets this singularity requirement. However it provides only a generic estimate of radiation burden that is technically defined only for a single reference phantom [7]. Effective dose ignores patient gender, size and age, factors that can markedly impact radiation risk and are of significant relevance to the pediatric population. Moreover, although characterizing radiation burden in terms of dose (CTDIvol, SSDE, organ dose, or

effective dose) is most common, what justifies and makes relevant this quantification is not the radiation dose itself but radiation risk. If there is no risk, there is no point in any dose quantification. It is possible, however, to combine the knowledge of organ dose and many radiobiologic factors that are of known relevance (e.g., age and gender) into a scalar quantity that can serve as an index of risk for optimization and minimization purposes [8, 9]. However such an undertaking requires the knowledge of the organ dose. Organ dose not only serves as a solid basis for any risk estimation but also enables dose comparisons across modalities. Therefore estimating patient-specific organ dose remains the cornerstone objective of patient-specific dosimetry. Precise estimation of organ dose is hampered by three significant challenges. First, it requires a reasonable knowledge of the patient anatomy, specifically where the organs are located. This is difficult because patients, particularly pediatric ones, come with a high degree of variability of anatomical construct and body habitus. Second, precise estimation of organ dose requires the knowledge of the X-ray irradiation condition as that changes from examination to examination. This is particularly challenging because modern CT scanners employ an adaptive irradiation condition (known as tube current modulation or TCM) that alters the X-ray flux over the patient habitus. Third, the patient anatomy and TCM need to be integrated so that one can determine how the radiation field encompasses an organ of interest, a necessary requirement to estimate the organ dose. In this paper, we demonstrate a framework through which these challenges can be overcome to enable a reasonably accurate estimation of organ dose.

S462

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

Organ dose estimation

Modeling irradiation condition

Organ dose is mainly governed by two key factors: patient anatomy (patient size, organ location, and tissue composition) and scanner output (influenced by X-ray tube voltage, bowtie filter, pitch, and tube current modulation). Methods to model each factor are detailed below.

With a representative model of patient anatomy, the next step for organ dose estimation is to model the irradiation condition. This process generally includes the detailed modeling of the CT system geometry, the X-ray tube motion, the bowtie filter, kVp and mA. Estimating organ dose under fixed tube current conditions is straightforward. This has been performed using a benchmarked Monte Carlo radiation transport program across the population of human models for a variety of CT systems [8, 16, 17]. With a fixed irradiation condition over the scanned field of view, organ doses are estimated by tallying the energy deposited in each organ. The organ doses for each CT protocol are normalized by the CTDIvol of the examination into so-called h values for each organ and each CT protocol. This normalization can also be done in terms of the dose length product (DLP) [18]. However, because h values are already specific to each protocol with a specific axial coverage, the values implicitly account for the scan length. The organ doses estimated with a fixed tube current cannot be directly applied to scans acquired with tube current modulation. TCM aims to achieve consistent image quality by dynamically adjusting the tube current based on patient attenuation. As the radiation field changes across the imaging volume, different organs receive different exposures. The field encompassing an organ further changes from patient to patient. Thus the knowledge of current modulation is necessary to determine the organ radiation field. The TCM profile can be directly extracted from the CT system if that information is made available by the manufacturer. Alternatively, it can be modeled based on patient attenuation. Different manufacturers use different modulation strategies; modeling the TCM should reflect that variability. A recent study developed a manufacturer-generalizable technique to model the TCM profile using a ray-tracing algorithm [19]. The algorithm determines the tube current based on the patient attenuation at a specific view, taking into account the geometry of the CT system, the poly-energetic X-ray energy spectrum, and the attenuation through the bowtie filter and the patient. The different modulation techniques are modeled using a modulation strength factor as shown in Fig. 3. With a reasonable knowledge of the irradiation field, one can match the field with the patient anatomy to estimate the organ dose values.

Modeling patient anatomy The first requirement for organ dose estimation is an adequately accurate knowledge of the patient anatomy. This goal can be approached with the use of computational human models. Among the efforts toward the development of such models [10–12], the work pursued by Duke University offers one of the most comprehensive in terms of scope and representation. This group has generated the largest database of human models covering a broad range of patient body habitus and reflecting the anatomical variability across the population into an atlas within which any given patient can find a reasonable match (Fig. 1). Each computational model is generated based on a clinical CT case [13, 14]. An initial model is created by segmenting the bones and major organs within the CT image volume. A 3-D surface is then fit to the polygon models using NURBS modeling software (Rhinoceros; McNeel North America, Seattle, WA). Other organs and structures are defined by morphing structures from existing male or female full-body adult and pediatric models. The volumes of the morphed organs and structures are checked and scaled, if necessary, to match ageinterpolated organ volume and anthropometry data. The full-body patient models include most of the radiosensitive organs defined by the International Commission on Radiological Protection’s publication 103 [15] and can be readily incorporated into simulation programs for image quality or dose estimation. There is a broad representation of human anatomy across the population, so a new clinical patient can be matched to a corresponding computational model that closely resembles the patient in terms of location of major organs. Such matching aims to achieve a reasonable approximation of patient anatomy. We have found trunk height to be a good first-order indicator to enable this matching. The trunk height can either be directly measured or estimated from the localizer image. Assuming the model in the atlas to be a hypothetical patient, four pairs of matched models are shown in Fig. 2 (male and female adults at 25% and 75% height and weight).

Organ dose equation With effective methods to approximate the patient anatomy and radiation field, dose to each organ of a patient undergoing

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

S463

Fig. 1 An example of 3-D frontal views of patient models developed by Duke University [12] (a), along with their body mass index (BMI) and age distribution (b). Each data-point in (b) represents a patient case

converted into a whole-body human model. The bands represent weight percentiles across ages

a CT examination with or without TCM, Dorgan, can be estimated [19] as

where mAz is the mA value at location z along the scan, N is the number of organ voxels in the axial slice at location z, CTDIvol is the CTDIvol of the exam, and horgan is the CTDIvolnormalized organ dose at a fixed mA described above. In the above equation, the first term is an organ-specific mA scaling factor computed from the TCM mA profile reflecting the strength of radiation field for the organ. For a protocol with a fixed current, this term equals unity.

X Dorgan ¼

mAz  N z

z∈forgang

X

z∈forgang

Nz

 CTDI vol  horgan ;

ð1Þ

S464

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

Fig. 2 An example of patient–model matching pairs as determined by trunk height are shown in (a) a 25th-percentile man, (b) a 75th-percentile man, (c) a 25th-percentile woman and (d) a 75th-percentile woman. In this example, the model on the left was taken as the patient and the one on the right was its closest match in terms of trunk height from the atlas.

Matches based on the trunk height provided a statistically accurate representation of the axial location of the organs, which can then be used in combination with the tube current modulation profile and the patient thickness to prospectively estimate organ doses in the scanned patient

Validation

method, Eq. 1; (4) estimated organ doses were compared against the Monte Carlo simulated organ dose values to quantify estimation accuracy. Figure 4 illustrates the validation results. The histograms of prediction errors for four organs are plotted for each exam. The errors are normalized by CTDIvol so that the accuracy can be quantified across different exams. Assuming an average CTDIvol of 10 mGy, the absolute median error across all organs was 0.64 mGy (−0.21 and 0.97 for 25th and 75th percentiles, respectively). For organs within the image coverage, the estimation error generally increased with the increasing level of tube current modulation. This can be explained by the fact that this technique approximates the organ dose under the TCM scan using the organ dose coefficients estimated from fixed mAs. As such, the method is expected to produce less accurate results when the current is changing rapidly in the axial dimension.

In the previous section we illustrated the process to estimate organ dose for a clinical patient under a tube current modulation condition. To quantify the accuracy of this formalism, we recently performed a validation study involving 58 computational models and patients (age range 18–78 years, weight range 57–180 kg) undergoing clinical abdominopelvic scans [20]. Although this validation was done in the context of adult imaging, the same procedure can be applied to pediatric patients. The study involved four steps: (1) to establish a gold standard, the actual tube current modulation was incorporated into our validated Monte Carlo program and all organ doses were quantified across the 58 patient models as a function of modulation strength; (2) each computational model, regarded as a clinical patient, was matched to one of the remaining 57 computational models in the atlas using a leave-one-out strategy; (3) organ doses were estimated using the proposed

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

S465

Fig. 3 Graphs show tube current modulation profiles at four modulation strengths. Each profile shows how the mA is modulated along the axial length of the patient. The higher-frequency oscillations reflect the changes in the mA as the tube rotates around the patient and encounters the thick (lateral) and thin (anterior–posterior) trajectories through the elliptical cross-section of the body. The parameter “a” indicates the

modulation strength, a gain factor dictating the adaptation of the mA to the patient thickness. a=0 indicates no modulation (fixed mA) while a=1 indicates the maximum mA modulation to fully compensate for the body thickness, aiming to achieve constant image noise regardless of the attenuation the X-rays encounter as they project through the body

Conclusion

range of patient anatomical diversity. The modeling of tube current modulation can be achieved by a simulation program that can represent TCM profiles across scanners. Depending on the specific requirements of clinical practice, organ dose estimation can be performed either prospectively or retrospectively. Prospective estimation can serve as the basis for the design of individualized protocols in relation to a targeted level of image quality. Retrospective estimation enables improved dose monitoring. It may further aid in the design of follow-up scans for radiosensitive patients and in effective communication with the patient pertaining to the justifiability and the optimality of the examination.

Characterizing the level of CT radiation for an individual patient has emerged as an unavoidable requirement to practice medical imaging. Among various dose metrics, organ dose is generally regarded as one of the best metrics to quantify individual radiation burden. It further serves as the basis for radiation risk estimation. To be accurate, the estimation of organ dose requires detailed modeling of the patient anatomy and the scanner irradiation condition, especially the tube current modulation profile. The modeling of patient anatomy can be achieved by an atlas-based strategy that covers a broad

S466

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467

Fig. 4 Histogram depicts error in predicting organ dose for the abdominopelvic scans in the (a) stomach, (b) liver, (c) kidney and (d) lung. The x-axis is the difference between the prospectively estimated organ doses (by patient matching and the application of Eq. 1) and the

actual organ doses retrospectively assessed by full Monte Carlo estimation on the patient data. The errors are normalized by the CTDIvol of the exam to make the error figure independent of the total dose that the patient received

Conflicts of interest Dr. Samei receives grant support for research from GE Healthcare, Siemens Healthcare and Carestream Health. Drs. Tian and Segars have no financial interests, investigational or off-label uses to disclose.

2. Brenner DJ, Hall EJ (2007) Computed tomography — an increasing source of radiation exposure. N Engl J Med 357:2277–2284 3. McCollough CH, Chen GH, Kalender W et al (2012) Achieving routine submillisievert CT scanning: report from the summit on management of radiation dose in CT. Radiology 264:567–580 4. Boone JM, Hendee WR, McNitt-Gray MF et al (2011) Radiation exposure from CT scans: how to close our knowledge gaps, monitor and safeguard exposure — proceedings and recommendations of the radiation dose summit, sponsored by NIBIB, February 24–25. Radiology 265:544–554

References 1. McCollough CH, Guimaraes L, Fletcher JG (2009) In defense of body CT. AJR Am J Roentgenol 193:28–39

Pediatr Radiol (2014) 44 (Suppl 3):S460–S467 5. McCollough CH, Leng S, Yu L et al (2011) CT dose index and patient dose: they are not the same thing. Radiology 259:311–316 6. Gold JE, Park JS, Punnett L (2006) Work routinization and implications for ergonomic exposure assessment. Ergonomics 49:12–27 7. Turner AC, Zankl M, DeMarco JJ et al (2010) The feasibility of a scanner-independent technique to estimate organ dose from MDCT scans: using CTDIvol to account for differences between scanners. Med Phys 37:1816–1825 8. Tian X, Li X, Segars WP et al (2013) Dose coefficients in pediatric and adult abdominopelvic CT based on 100 patient models. Phys Med Biol 58:8755–8768 9. Li X, Samei E, Segars WP et al (2011) Patient-specific radiation dose and cancer risk estimation in CT: part II. Application to patients. Med Phys 38:408–419 10. Petoussi-Henss N, Zanki M, Fill U et al (2002) The GSF family of voxel phantoms. Phys Med Biol 47:89–106 11. Lee C, Lodwick D, Hurtado J et al (2010) The UF family of reference hybrid phantoms for computational radiation dosimetry. Phys Med Biol 55:339–363 12. Segars WP, Bond J, Frush J et al (2013) Population of anatomically variable 4D XCAT adult phantoms for imaging research and optimization. Med Phys 40:043701

S467 13. Segars WP, Sturgeon G, Mendonca S et al (2010) 4D XCAT phantom for multimodality imaging research. Med Phys 37: 4902–4915 14. Norris H, Zhang Y, Bond J et al (2014) A set of 4D pediatric XCAT reference phantoms for multimodality research. Med Phys 41:033701 15. International Commission on Radiological Protection (2007) The 2007 recommendations of the international commission on radiological protection, ICRP publication 103. Ann ICRP 37:2–4 16. Tian X, Li X, Segars WP et al (2014) Pediatric chest and abdominopelvic CT: organ dose estimation based on 42 patient models. Radiology 270:535–547 17. Li X, Samei E, Segars WP et al (2011) Patient-specific radiation dose and cancer risk for pediatric chest CT. Radiology 259:862–874 18. Dixon RL, Boone JM (2013) Dose equations for tube current modulation in CT scanning and the interpretation of the associated CTDIvol. Med Phys 40:111920 19. Li X, Segars WP, Samei E (2014) Generic assessment of organ dose for CT scans with different tube current modulation schemes. Phys Med Biol [In press] 20. Tian X, Li X, Segars WP et al (2014) Prospective optimization of CT under tube current modulation. I: Organ dose. SPIE Medical Imaging, International Society for Optics and Photonics, p 90331Q

Determining organ dose: the holy grail.

Among the various metrics to quantify CT radiation dose, organ dose is generally regarded as one of the best to reflect patient radiation burden. Orga...
11MB Sizes 0 Downloads 5 Views