|Year : 2013 | Volume
| Issue : 3 | Page : 430-435
Dosimetric impact of Acuros XB dose calculation algorithm in prostate cancer treatment using RapidArc
Suresh Rana1, Kevin Rogers2, Terry Lee2, Daniel Reed2, Christopher Biggs2
1 Department of Radiation Oncology, Arizona Center for Cancer Care, Peoria, Arizona; Department of Medical Physics, ProCure Proton Therapy Center, Oklahoma City, OK, USA
2 Department of Radiation Oncology, Arizona Center for Cancer Care, Peoria, Arizona, USA
|Date of Web Publication||8-Oct-2013|
Department of Medical Physics, ProCure Proton Therapy Center, 5901 West Memorial Road, Oklahoma City, OK 73142
Source of Support: None, Conflict of Interest: None
Purpose: The purpose of this study is to assess the dosimetric impact of Acuros XB dose calculation algorithm (AXB), in comparisons with Anisotropic Analytical Algorithm (AAA) calculations in prostate cancer treatment using RapidArc.
Materials and Methods: A computed tomography (CT) dataset of low-risk prostate cancer patients treated at Arizona Center for Cancer Care was selected and contoured for prostate, seminal vesicles, and organs at risk (OARs)(rectum, bladder, and femur heads). Plans were created for 6 MV photon beam using RapidArc technique in Eclipse treatment planning system. Dose calculations were performed with AAA and AXB for same number of monitor units and identical beam setup. Mean and maximum doses to planning target volume (PTV) and OARs were analyzed. Additionally, minimum dose to PTV and V100 was analyzed. Finally, point-dose difference between planar dose distributions of AAA and AXB plans was investigated.
Results: The highest dose difference was up to 0.43% (range: 0.05−0.43%, P> 0.05) for PTV and 1.98% (range: 0.22−1.98%, P> 0.05) for OARs with AAA predicting higher dose than AXB. The V100 values of AAA plans (95 %) and AXB plans (range: 93.1−97.9 %) had an average difference of 0.89±1.47% with no statistical significance (P = 0.25411). The point-dose difference analysis showed that AAA predicted higher dose than AXB at significantly higher percentage (in average 94.15) of total evaluated points.
Conclusion: The dosimetric results of this study suggest that the AXB can perform the dose computation comparable to AAA in RapidArc prostate cancer treatment plans that are generated by a partial single-arc technique.
Keywords: Anisotropic analytical algorithm, Acuros, prostate cancer, RapidArc
|How to cite this article:|
Rana S, Rogers K, Lee T, Reed D, Biggs C. Dosimetric impact of Acuros XB dose calculation algorithm in prostate cancer treatment using RapidArc. J Can Res Ther 2013;9:430-5
|How to cite this URL:|
Rana S, Rogers K, Lee T, Reed D, Biggs C. Dosimetric impact of Acuros XB dose calculation algorithm in prostate cancer treatment using RapidArc. J Can Res Ther [serial online] 2013 [cited 2020 Jan 24];9:430-5. Available from: http://www.cancerjournal.net/text.asp?2013/9/3/430/119328
| > Introduction|| |
Prostate cancer is the second most commonly diagnosed cancer among men in America after skin cancer.  The external beam radiation therapy (EBRT) has played an important role in the treatment of prostate cancer over the years. Dose escalation creates problem with normal tissue tolerance, however, with intensity modulation radiation therapy (IMRT) normal tissues can be spared. , The recent development of volumetric modulated arc therapy (VMAT) technique has enabled the delivery of conformal dose distribution to the target while minimizing the dose to the critical structures. ,, RapidArc (Varian Medical Systems, Palo Alto, CA) is one of such VMAT techniques that delivers modulated radiation beams with simultaneous adjustment of multi leaf collimator (MLC) field aperture, dose rate, and gantry rotation speed. ,
In order to achieve the therapeutic advantage from RapidArc technique for the prostate cancer, more precise dose calculation using treatment planning system (TPS) is essential. The American Association of Physicists in Medicine (AAPM) has recommended that the uncertainty in the computed dose distribution to be less than 2%.  This constraint on the dose computation may be achieved for commercially existing dose calculation algorithms when considering a homogenous tissue. However, significant errors in the dose computations may result in the heterogeneous media such as human body where electron densities of tissues deviate from water. ,, Thus, tissue heterogeneities in the photon beam path must be incorporated in the dose calculation algorithm for achieving accurate RapidArc planning. The introduction of convolution-superposition algorithms such as Anisotropic Analytical Algorithm (AAA) in the Eclipse TPS (Varian Medical Systems, Palo Alto, CA) that better account for the heterogeneity has improved the estimation of dose distribution ,, in radiation treatment plans. More recently, the new commercially available Acuros XB advanced dose calculation algorithm (AXB) was implemented within the Eclipse TPS at Arizona Center for Cancer Care.
The AXB, first published by Vassiliev et al.,  is considered to be similar to classic Monte Carlo methods for accurate modeling of dose deposition in heterogeneous media. ,, The AXB has been described in details in previous publications ,,, and a brief description on AXB is provided here. The AXB utilizes the Linear Boltzmann Transport Equation (LBTE) and solves numerically that describes the macroscopic behavior of radiation transport as they travel through and interact with the matter. The AXB implementation in the Eclipse TPS has two models. The first one is the multiple source model already implemented for the AAA. , The second one is the radiation transport model, which allows calculating the dose accounting for the elemental composition of specific anatomical regions as derived by the Computed Tomography (CT) dataset. ,,,
By contrast, the AAA, first developed by Tillikainen et al.,, and Ulmer et al.,,, is an analytical photon dose calculation algorithm based on a pencil-beam convolution-superposition technique. The pencil beams in the AAA include separately modeled contributions from primary photons, extra-focal photons, and contaminating electrons. The total energy deposited by each beam is obtained by the convolution of the separately-modeled contributions of three different photon sources and final dose is calculated by the superposition of the contributions from the beams.  The tissue heterogeneity in the AAA is handled by scaling of primary photons and photon scatter kernel scaling in lateral directions according to local electron density. , For detailed description on the AAA, readers are advised to refer to work by Tillikainen et al. ,
Several recent studies have reported the use of RapidArc planning for the prostate cancer. ,,,,,, These studies utilized the convolution-superposition algorithms for the dose computation, and clinical validation of AXB is yet to be done for the RapidArc prostate cancer treatment plans. The purpose of this study is to assess the clinical dosimetric impact of AXB, in comparisons with well known and validated AAA calculations, ,,,,, for ten different RapidArc prostate cancer treatment plans that were generated using a partial single-arc technique.
| > Materials and Methods|| |
Computed tomography simulation and contouring
Ten patients with a localized prostate cancer (T0 and T1 Stages) treated with EBRT at Arizona Center for Cancer Care were selected for this study. All ten patients were immobilized in a supine position on a flat tabletop of General Electric LightSpeed CT Scanner using Head Sponge and Vac-Lok Cushions (CIVCO Medical Solutions, Kalona, Iowa). Patients underwent standard CT simulation. The CT scans were acquired with 512×512 pixels at 0.25 cm slice spacing from the top of the iliac crests superiorly to the perineum inferiorly. The clinical target volume (CTV) comprised of prostate and seminal vesicles were contoured by the radiation oncologist on the axial CT slices. The rectum, bladder, and femur heads as organs at risk (OARs) were delineated based on the axial CT images. The planning target volume (PTV) was generated from the CTV by a uniform expansion of 5 mm in all directions [Figure 1] based on the recommendation of Radiation Therapy Oncology Group (RTOG) 0815. The 5 mm PTV margin was essential to compensate for the variability of treatment setup and internal organ motion.
|Figure 1: An axial computed tomorgraphy slice of a prostate cancer patient. The PTV was generated from the CTV by a uniform expansion of 5 mm in all directions. PTV: Planning Target Volume, CTV: Clinical Target Volume|
Click here to view
RapidArc planning and optimization
The beam parameters for RapidArc plans were setup in the Eclipse TPS using Varian Clinac iX (Varian Medical Systems, Palo Alto, CA) with 120-leaf millennium MLC and 6 MV X-rays at a dose rate of 600 MU/min was used. For a partial single-arc technique, a single gantry rotation in a clockwise direction (Varian IEC scale) was used. The collimator angle was set at 5 0 for all plans. Some variations in the range of gantry angles (223 ± 7 → 135 ± 5 degrees) were allowed to avoid the direct beam entrance through Varian couch-rails [Figure 2]. The Beam's-Eye-View graphics in the Eclipse TPS was used to better define the field sizes of the coplanar arc. The isocenter of the plans was placed at the center of the prostate. All plans were generated using Progressive Resolution Optimizer (PRO) (version 10.0.26) in the Eclipse TPS, and the volumetric dose optimization method followed the same systematic strategy regarding the objectives and priorities. The total dose prescribed to the PTV was 79.2 Gy with a daily dose of 1.8 Gy to be treated in 44 fractions, and this treatment regimen was applied for all ten patients in this study. The dose constraints to the OARs are summarized in [Table 1]. These dose constraints were determined based on RTOG -0815 guidelines.
|Figure 2: A transversal view of RapidArc plan setup using partial single-arc technique|
Click here to view
Dose calculation and plan normalization
All optimized RapidArc prostate cancer treatment plans were computed with two dose calculation algorithms implemented in the Eclipse TPS: (1) AAA, version 10.0.26; and (2) AXB, version 10.0.26. The dose calculation grid was set to 2.5 mm for all cases.
The dose-to-water calculations were computed by AAA and the calculated plans are referred as AAA plans. The AAA plans were then normalized such that the 100% of the prescribed dose covered the 95% of the PTV, and the monitor units (MUs) from the normalized AAA plans were recorded.
The option to calculate either dose-to-medium or dose-to-water is available in AXB. However, for this study, dose-to-medium calculations were performed and the plans computed by the AXB are referred as AXB plans. For the dose-to-medium calculations in AXB, the macroscopic energy deposition cross-section and atomic density are based on the material properties of local voxel. , In contrast, the energy deposition cross-sections for water are used in place of those for the local media in the case of the dose-to-water calculations. ,, The dose calculations in the AXB plans were done using identical beam setup and same number of MUs as in the normalized AAA plans.
Plan evaluation and statistical analysis
The cumulative dose-volume histograms (DVH) of all calculated RapidArc prostate cancer treatment plans (AAA and AXB) were generated in the Eclipse TPS for subsequent dosimetric analysis. The maximum and mean dose to the PTV and OARs were evaluated. Additionally, the minimum dose to the PTV and the percentage of PTV covered by 100% of the prescribed dose (V100) was investigated. The average cumulative DVH of the OARs over all ten patients was created. Finally, the planar dose distributions of the calculated plans were exported from the Eclipse TPS and MapCHECK software, version 5.00.00 (Sun Nuclear, Melbourne, FL) was used to perform the absolute point-dose comparisons. In order to test the observed differences for the evaluated parameters, statistical analysis was done using two-sided student's t-test, and P value of less than 0.05 (i.e., P< 0.05) was considered to be statistically significant.
| > Results|| |
In this study, we analyzed the twenty RapidArc prostate cancer treatment plans of ten patients. The clinical dosimetric impact of AXB, in comparisons with AAA calculations, was assessed.
[Table 2] summarizes the doses to the PTV and OARs for the selected DVH parameters and the values are averaged over the ten analyzed patients. The AAA predicted higher minimum, mean and maximum doses to the PTV but the dose difference was less than 0.5% (range: 0.05−0.43%) and did not show the statistical significance (P > 0.05). For OARs, the maximum doses in the AAA plans were higher by in average 0.58 % (rectum: 0.22%, bladder: 0.64%, and femur heads: 0.88 %) when compared against the AXB plans. This trend continued with AAA predicting higher mean dose to the OARs by in average 1.44% (rectum: 1.98 %, bladder: 1.15 %, and femur heads: 1.18 %). No statistical significant differences (P > 0.05) were observed for the critical structures.
[Figure 3] shows the percentage of PTV covered by 100% of the prescribed dose (V100) and the V100 values are averaged over the ten analyzed patients. The V100 values of AAA plans (95 %) and AXB plans (range: 93.1−97.9 %) had an average difference less than 1% and the difference did not show the statistical significance (P = 0.25411).
|Figure 3: PTV coverage by 100 % of the prescribed dose of 79.2 Gy in AAA and AXB plans.PTV: Planning Target Volume, V100: Percent of PTV covered by 100% of the prescribed dose, AAA: Anisotropic Analytical Algorithm, AXB: Acuros XB Algorithm(Values are averaged over the 10 analyzed patients and error bars represent the standard deviation)|
Click here to view
Point-dose difference analysis
[Figure 4] is the histogram that shows the percent of points versus the percent point-dose differences between two calculated plans (AAA and AXB) for absolute dose comparison and the values are averaged over the ten analyzed patients. In [Figure 4], the red column bars represent the average percent of hot points that have the positive point-dose difference (AAA's prediction higher than AXB), whereas, the green column bars represent the average percent of cold points that have the negative point-dose difference (AXB's prediction higher than AAA). The average ratio of hot points (94.15%) to cold points (5.85%) was about 16 and this ratio suggests that the AAA's dose prediction was higher at the majority of the points.
|Figure 4: An absolute point-dose comparison between AAA and AXB plans. Red column bar represents the average percent of hot points (AAA's prediction higher than AXB) and green column bar represents the average percent of cold points (AXB's prediction higher than AAA) Point-Dose Difference: (AAA-AXB)/AAA. AAA: Anisotropic Analytical Algorithm, AXB: Acuros XB Algorithm(Values are averaged over the 10 analyzed patients and error bars represent the standard deviation)|
Click here to view
| > Discussion|| |
Dose calculations present challenges especially when photon beam travels through tissues with different densities before reaching the target.  The accurate modeling of primary beam attenuation and lateral scatter due to the presence of different media heterogeneities along the photon beam path is essential to avoid the dose overestimation or underestimation. Recently, the accuracy of AXB, which is similar to Monte Carlo method, ,, has been investigated in heterogeneous media ,,, and the results were found to be almost identical to the Monte Carlo simulations.
In this study, the highest dose difference was up to 0.43% for the PTV and 1.98% for the critical structures with AAA predicting higher dose than AXB. Although, we could not find the literature to make the direct comparisons against our data for RapidArc prostate cancer treatment plans, it is relevant to mention two clinical studies , done by Fogliata et al., on AAA and AXB. In the first study,  Fogliata et al., reported that the AAA predicted higher dose than AXB and the difference between AAA and AXB in muscle tissue was in average 1.6% (3D conformal therapy, 6 MV photon beam). In the second study,  where target is in the soft tissue, the mean PTV dose was found to be lower for Acuros XB, with a range of 0.4% ± 0.6% (IMRT) and 1.3 % ± 0.2% (RapidArc) for 6 MV photon beam. In the same study,  the mean doses to OAR differences up to 3% of the mean structure dose in the worst case with AAA predicting higher dose.
It is evident that the percent dose differences between AAA and AXB calculations among our study and Fogliata's findings are different. The AXB can calculate the dose-to-medium by distinguishing between different types of soft tissues characterized by significantly different chemical composition. ,,, However, those characteristics are not available in the AAA, where the dose calculation accounts only for the different densities of the materials and the dose is computed as dose to density rescaled water. ,,,,, Thus, the variations in disagreement values (e.g. percent dose differences) between the AAA and AXB calculations were expected for different types of tissues (e.g. femur heads, muscle, heart, and spinal cord). However, the published results from clinical cases , and the dosimetric results of this study demonstrate that AXB is capable of modeling radiotherapy dose deposition similar to that of AAA. To further illustrate the AXB's ability to compute lower dose to critical structures compared to AAA, the average cumulative DVHs over all ten patients are presented in [Figure 5].
|Figure 5: The cumulative dose-volume histogram of rectum, bladder and femur heads from RapidArc prostate cancer treatment plans (AAA: Dotted line and AXB: Solid line) averaged over the 10 analyzed patients. AAA: Anisotropic Analytical Algorithm, AXB: Acuros XB Algorithm|
Click here to view
In [Figure 3], the average PTV coverage for the 100% prescribed dose was slightly inferior in the AXB plans than that in the AAA plans and this clearly indicates that the required PTV coverage (V100 ≥ 95% of PTV) at our clinic was not achieved for some of the AXB plans. However, the V100 values of AAA and AXB were differed by an average of 0.89%, which is a smaller value compared to its standard deviation (±1.47%).
Point-dose difference analysis
The results obtained from the comparisons of absolute planar dose distributions showed that the AAA predicted higher dose than AXB at significantly higher percentage (in average 94.15) of the total points. Although the majority of the points (in average 95.82% of total points) had the dose differences within ±2%, the point-dose discrepancies between AAA and AXB plans were not consistent. In the future, we aim to continue this study using the pelvis-prostate phantom and taking measurements with ionization chamber and radiochromic films. This will allow us to determine the agreement between the calculated (AAA and AXB) and measured values in the low and high dose regions. Furthermore, it is very essential to investigate the point-dose differences in the high dose gradients, which have the potential to cause larger dose errors in the associated point dose measurements, because very small positional errors can cause large dose differences. While our clinical dosimetric study on RapidArc prostate cancer treatment plans demonstrated the AXB's capability of computing the dose comparable to AAA, further study is needed for prostate patients with prosthesis, which has significantly higher electron density value than that of soft tissue.
| > Conclusion|| |
The dosimetric results of this study suggest that the AXB can perform the dose computation comparable to AAA in RapidArc prostate cancer treatment plans that are generated by a partial single-arc technique.
| > Acknowledgment|| |
The authors would like to thank Dr. Indra J. Das, Vice Chair, Professor, and Director of Medical Physics at Indiana University School of Medicine for providing valuable suggestions during the preparation of this manuscript.
| > References|| |
|1.||Howlader N, Noone AM, Krapcho M, Neyman N, Aminou R, Altekruse SF, et al.SEER Cancer Statistics Review. Bethesda, MD: National Cancer Institute; 2009. |
|2.||Yoo S, Wu QJ, Lee WR, Yin FF. Radiotherapy treatment plans with RapidArc for prostate cancer involving seminal vesicles and lymph nodes. Int J Radiat Oncol Biol Phys 2010;76:935-42. |
|3.||Otto K. Volumetric modulated arc therapy: IMRT in a single gantry arc. Med Phys 2008;35:310-7. |
|4.||Palma D, Vollans E, James K, Nakano S, Moiseenko V, Shaffer R, et al. Volumetric modulated arc therapy for delivery of prostate radiotherapy: Comparison with intensity-modulated radiotherapy and three-dimensional conformal radiotherapy. Int J Radiat Oncol Biol Phys 2008;72:996-1001. |
|5.||Ling CC, Zhang P, Archambault Y, Bocanek J, Tang G, Losasso T. Commissioning and quality assurance of RapidArc radiotherapy delivery system. Int J Radiat Oncol Biol Phys 2008;72:575-81. |
|6.||Tang G, Earl MA, Luan S, Wang C, Mohiuddin MM, Yu CX. Comparing radiation treatments using intensity-modulated beams, multiple arcs, and single arcs. Int J Radiat Oncol Biol Phys 2010;76:1554-62. |
|7.||Papanikolaou N, Battista JJ, Boyer AL, Kappas C, Klein C, Mackie TR, et al. Tissue inhomogeneity corrections for megavoltage photon beams. Report of Task Group No. 65 of the Radiation Therapy Committee of the American Association of Physicist in Medicine, AAPM, 2004; Report No.:85. |
|8.||Breitman K, Rathee S, Newcomb C, Murray B, Robinson D, Field C, et al. Experimental validation of the Eclipse AAA algorithm. J Appl Clin Med Phys 2007;8:76-92. |
|9.||Rønde HS, Hoffmann L. Validation of Varian's AAA algorithm with focus on lung treatments. Acta Oncol 2009;48:209-15. |
|10.||Van Esch A, Tillikainen L, Pyykkonen J, Tenhunen M, Helminen H, Siljamäki S, et al.Testing of the analytical anisotropic algorithm for photon dose calculation. Med Phys 2006;33:4130-48. |
|11.||Vassiliev O, Wareing T, McGhee J, Failla G, Salehpour M, Mourtada F. Validation of a new grid based Blotzmann equation solver for dose calculation in radiotherapy with photon beams. Phys Med Biol 2010;55:581-98. |
|12.||Bush K, Gagne IM, Zavgorodni S, Ansbacher W, Beckham W. Dosimetric validation of Acuros XB with Monte Carlo methods for photon dose calculations. Med Phys 2011;38:2208-21. |
|13.||Fogliata A, Nicolini G, Clivio A, Vanetti E, Mancosu P, Cozzi L. Dosimetric validation of Acuros XB Advanced Dose Calculation algorithm: Fundamental characterization in water. Phys Med Biol 2011;56:1879-904. |
|14.||Vassiliev ON, Wareing TA, Davis IM, McGhee J, Barnett D, Horton JL, et al. Feasibility of a multigroup deterministic solution method for three-dimensional radiotherapy dose calculations. Int J Radiat Oncol Biol Phys 2008;72:220-7. |
|15.||Tillikainen L, Helminen H, Torsti T, Siljamäki S, Alakuijala J, Pyyry J, et al. 3D pencil-beam-based superposition algorithm for photon dose calculation in heterogeneous media. Phys Med Biol 2008;53:3821-39. |
|16.||Tillikainen L, Siljamäki S, Helminen H, Alakuijala J, Pyyry J. Determination of parameters for a multiple-source model of megavoltage photon beams using optimization methods. Phys Med Biol 2007;52:1441-67. |
|17.||Ulmer W, Harder D. A triple gaussian pencil beam model for photon beam treatment planning. Z Med Phys 1995;5:25-30. |
|18.||Ulmer W, Harder D. Applications of a triple gaussian pencil beam model for photon beam treatment planning. Z Med Phys 1996;6:68-74. |
|19.||Ulmer W, Kaissl W. The inverse problem of a gaussian convolution and its application to the finite size of the measurement chambers/detectors in photon and proton dosimetry. Phys Med Biol 2003;48:707-27. |
|20.||Robinson D. Inhomogeneity correction and the analytic anisotropic algorithm. J Appl Clin Med Phys 2008;9:112-22. |
|21.||Kjaer-Kristoffersen F, Ohlhues L, Medin J, Korreman S. RapidArc volumetric modulated therapy planning for prostate cancer patients. Acta Oncol 2009;48:227-32. |
|22.||Wolff D, Stieler F, Welzel G, Lorenz F, Abo-Madyan Y, Mai S, et al. Volumetric modulated arc therapy (VMAT) vs. serial tomotherapy, step-and-shoot IMRT and 3D-conformal RT for treatment of prostate cancer. Radiother Oncol 2009;93:226-33. |
|23.||Guckenberger M, Richter A, Krieger T, Wilbert J, Baier K, Flentje M. Is a single arc sufficient in volumetric-modulated arc therapy (VMAT) for complex-shaped target volumes? Radiother Oncol 2009;93:259-65. |
|24.||Shaffer R, Morris WJ, Moiseenko V, Welsh M, Crumley C, Nakano S, et al. Volumetric modulated arc therapy and conventional intensity-modulated radiotherapy for simultaneous maximal intraprostatic boost: A planning comparison study. Clin Oncol (RCollRadiol) 2009;21:401-7. |
|25.||Fogliata A, Nicolini G, Clivio A, Vanetti E, Cozzi L. Dosimetric evaluation of Acuros XB Advanced Dose Calculation algorithm in heterogeneous media. Radiat Oncol 2011;6:82. |
|26.||Fogliata A, Nicolini G, Clivio A, Vanetti E,Cozzi L. On the dosimetric impact of inhomogeneity management in the Acuros XB algorithm for breast treatment. Radiat Oncol 2011;6:103. |
|27.||Fogliata A, Nicolini G, Clivio A, Vanetti E, Cozzi L. Critical appraisal of Acuros XB and Anisotropic Analytic Algorithm dose calculation in advanced non-small-cell lung cancer treatments. Int J Radiat Oncol Biol Phys 2012;83:1587-95. |
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2]
|This article has been cited by|
||Proton Therapy vs. VMAT for Prostate Cancer: A Treatment Planning Study
| ||Suresh Rana,ChihYao Cheng,Yuanshui Zheng,Dina Risalvato,Nancy Cersonsky,Eric Ramirez,Li Zhao,Gary Larson,Carlos Vargas |
| ||International Journal of Particle Therapy. 2014; 1(1): 22 |
|[Pubmed] | [DOI]|
||The accuracy of the Acuros XB algorithm in external beam radiotherapy – a comprehensive review
| ||Jarkko Ojala |
| ||International Journal of Cancer Therapy and Oncology. 2014; 2(4): 020417 |
|[Pubmed] | [DOI]|
||The quality assurance of volumetric modulated arc therapy (VMAT) plans for early stage prostate cancer: a technical note
| ||Craig A. Elith,Shane E. Dempsey,Fred Cao,Afrooz Farshadi,Helen M. Warren-Forward |
| ||Journal of Medical Radiation Sciences. 2014; : n/a |
|[Pubmed] | [DOI]|