- Email: [email protected]
Monte Carlo evaluation of Acuros XB dose calculation Algorithm for intensity modulated radiation therapy of nasopharyngeal carcinoma
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
Contents lists available at ScienceDirect
Radiation Physics and Chemistry journal homepage: www.elsevier.com/locate/radphyschem
Monte Carlo evaluation of Acuros XB dose calculation Algorithm for intensity modulated radiation therapy of nasopharyngeal carcinoma ⁎
Peter C.Y. Yeha,b, C.C. Leec,d, T.C. Chaoc,e, C.J. Tungc,e, a
Department of Radiation Oncology, Tungs MetroHarbor General Hospital, Taichung, Taiwan Jen-Teh Junior College of Medicine, Nursing and Management, Miaoli, Taiwan c Department of Medical Imaging and Radiological Sciences, College of Medicine, Chang Gung University, Taiwan d Department of Radiation Oncology, Chang Gung Memorial Hospital, Taiwan e Institute for Radiological Research, Chang Gung University/Chang Gung Memorial Hospital, Taiwan b
A R T I C L E I N F O
A BS T RAC T
Keywords: Acuros XB Monte Carlo Anisotropic analytical algorithm Nasopharyngeal carcinoma Intensity modulated radiation therapy
Intensity-modulated radiation therapy is an effective treatment modality for the nasopharyngeal carcinoma. One important aspect of this cancer treatment is the need to have an accurate dose algorithm dealing with the complex air/bone/tissue interface in the head-neck region to achieve the cure without radiation-induced toxicities. The Acuros XB algorithm explicitly solves the linear Boltzmann transport equation in voxelized volumes to account for the tissue heterogeneities such as lungs, bone, air, and soft tissues in the treatment field receiving radiotherapy. With the single beam setup in phantoms, this algorithm has already been demonstrated to achieve the comparable accuracy with Monte Carlo simulations. In the present study, five nasopharyngeal carcinoma patients treated with the intensity-modulated radiation therapy were examined for their dose distributions calculated using the Acuros XB in the planning target volume and the organ-at-risk. Corresponding results of Monte Carlo simulations were computed from the electronic portal image data and the BEAMnrc/DOSXYZnrc code. Analysis of dose distributions in terms of the clinical indices indicated that the Acuros XB was in comparable accuracy with Monte Carlo simulations and better than the anisotropic analytical algorithm for dose calculations in real patients.
1. Introduction Nasopharyngeal carcinoma (NPC) is a unique primary head and neck cancer arising from the nasopharynx with an estimated 86,700 new cases and 50,800 deaths in 2012 worldwide. High incidence rates are observed in southeastern China, Hong Kong, Malaysia, Indonesia and Singapore (Torre et al., 2015). Intensity-modulated radiation therapy (IMRT) with concurrent CDDP+5-FU chemotherapy is the primary treatment of choice for NPC because of its unique anatomical location and sensitivity to radiotherapy (Sun et al., 2014; Tao et al., 2015). IMRT provides the steep radiation dose gradient to produce a high degree of conformal tumor target coverage and a sparing of normal tissues. The accurate dose calculations are required to obtain therapeutic advantages of the IMRT. One important aspect of IMRT for NPC patients is the need to have an accurate dose calculation algorithm to deal with the effects on the complex air/bone/tissue interface for achieving the cure without radiation-induced toxicities. Although Monte Carlo (MC) calculations had the best agreement with measured data within the inhomogeneous
⁎
region (Ojala, 2014), these calculations require a large number of individual particles transporting in matter, thus resulting in an extensive computing time. This makes MC simulations virtually impractical in the clinical environment even with the variance reduction method, efficient sampling and coding technique (Bush et al., 2011). The anisotropic analytical algorithm (AAA), a convolution/ superposition method, has been widely utilized for dose calculations in the treatment planning system (TPS). However, AAA was reported to lead to dose discrepancies in the air/bone/tissue region due to its inherent pencil beam kernel and independent depth/lateral scaling of the kernel for heterogeneity correction (Kan et al., 2013a). The Acuros XB (AXB) algorithm provides a new advanced dose calculation method (Varian Medical Systems, Palo Alto, CA) applied in the TPS. It explicitly solves the linear Boltzmann transport equation by a deterministic method using discretized cross sections as radiations interact with the voxel volumes in matter. AXB makes use of the chemical composition of each material in the volume during radiation transport (Kan et al., 2013a, 2013b). Therefore, it directly accounts for the effects on tissue heterogeneities. With the single beam setup in
Correspondence to: Department of Medical Imaging and Radiological Sciences, Chang Gung University, 259 Wenhua 1st Rd., Guishan Dist., Taoyuan City 33302, Taiwan. E-mail addresses: [email protected], [email protected] (C.J. Tung).
http://dx.doi.org/10.1016/j.radphyschem.2017.02.025 Received 17 September 2016; Received in revised form 8 February 2017; Accepted 10 February 2017 0969-806X/ © 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Yeh, P.C.Y., Radiation Physics and Chemistry (2017), http://dx.doi.org/10.1016/j.radphyschem.2017.02.025
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
P.C.Y. Yeh et al.
phantoms, AXB has already been demonstrated to achieve the required accuracy compared with MC simulations but with faster computing speed and without statistical noise (Failla et al., 2010). For instance, Bush et al. (2011) compared AXB and MC calculated doses in a heterogeneous air/bone/lung phantom for 6 and 18 MV photon beams. They showed the maximum discrepancies were within 2% in lung, 1.8% in bone and 4.5% in air. Han et al. (2011) noted that the agreement between AXB and MC doses in a soft tissue/bone/lung phantom for 6 MV photon beam was within 2%. Although the accuracy of AXB has been shown in heterogeneous phantoms, the clinical validation of AXB in real patients treated with IMRT should be established, especially for tumors in the head and neck region where a complex air/bone/soft tissue interface could cause dose perturbations. The present study aims to compare dose distributions calculated using the AXB algorithm and the MC simulation and to evaluate the clinical impact of AXB on NPC patients treated with the IMRT. Five NPC patients were studied with measurement-based MC (MBMC) calculations applying the electronic portal image data and the BEAMnrc/DOSXYZnrc code. Analysis of dose distributions in the planning target volume (PTV) and the organs-at-risk (OAR) were made. It indicated that the AXB algorithm was in comparable accuracy with the Monte Carlo simulation in the head and neck region. 2. Materials and methods The AXB algorithm solves the linear Boltzmann transport equation of coupled photon and electron fluences in a voxel volume of matter through several steps. First, the external photon and electron sources are transported into the volume using ray-tracing techniques. Then, a finite-element method is used to find the energy- and angulardependent fluences in this volume by applying discretized cross sections and stopping powers. Since the AXB utilizes mass densities and atomic compositions in the interaction cross sections, its calculation is analogous to the MC simulation. Comparing to the stochastic MC simulation, however, the AXB greatly reduces the computing time because of its deterministic computation. In the present study, the AXB version 13.026 in the Eclipse TPS (Varian Medical Systems, Palo Alto, CA) was used with a grid size of 2 mm2. The AXB modeled four radiation sources: (1) primary source or bremsstrahlung photons created in the target, (2) extra focal source or photons resulting from interactions in the accelerator head (the flattening filter, primary collimators, and secondary jaws), (3) electron contamination, and (4) photons scattered from wedges (Failla et al., 2010). Five NPC patients treated with the IMRT received cumulative doses of 70 Gy in 35 fractions (Yeh et al., 2014). Seven coplanar 6 MV photon beams were used with the dynamic sliding window technique through a Millennium 120-leaf multi-leaf collimator (MLC) in the Varian 21EX linear accelerator. MC simulations were performed using the MBMC method, which applied the electronic portal image data and the BEAMnrc/DOSXYZnrc code version 2007 (Rogers et al., 1995). The MBMC method has been described in detail previously (Lin et al., 2009) and is summarized here with the aid of Fig. 1. First, an incident parallel circular electron beam with a Gaussian intensity distribution of full-width at half-maximum equal to 0.12 cm and a mean electron energy of 6.3 MeV with 3% energy spread was incident onto the Varian 21EX linear accelerator tungsten target to generate an open-field phase-space file, i.e. data on energies, positions, directions, and weightings of every particles crossing the scoring plane at 80 cm from the source. Then, MC simulations were performed by applying the variance reduction technique with the directional bremsstrahlung splitting. The splitting-field source-to-surface distance was set at 100 cm. The splitting-field radius was equal to the field size. The Russian roulette plane was chosen above the bottom of the flattening filter. Here at least 3.0×107 particles were simulated in each IMRT field to reduce the uncertainty to ≤2% in the phase-space file. Next, an efficiency map of each IMRT field was obtained from data
Fig. 1. The MBMC simulation setup.
collected on the amorphous silicon aS1000 electronic portal imaging device to adjust the weighting of each particle in the phase-space file. Further, the dose distribution in the patient was calculated using the DOSXYZ code (Rogers et al., 1995) and the efficiency map of IMRT photon beams transported through the patient. Finally, the calculated dose-to-medium could be converted to dose-to-water using the waterto-medium stopping power ratios (Siebers et al., 2000). To allow uncertainties in the patient positioning, alignment and respiratory motion during the IMRT, PTV was determined to be the irradiated tumor volume plus a 3–5 mm margin. The PTV dose distribution was evaluated by V > 95%, the percent PTV volume receiving ≥95% of the prescribed dose (70 Gy). The homogeneity index (HI) was evaluated by the ratio (D2%-D98%)/D50%, where D2%, D50% and D98% are the minimum doses received by 2%, 50% and 98% of the PTV volume, respectively. A lower HI indicated a better dose homogeneity or less cold/hot spots (Kan et al., 2013a). Although air cavities were usually included in the PTV, a common practice among radiation oncologists during the treatment planning, PTV, V > 95% and HI were all determined by either including or excluding air cavities in the target volume in order to assess the dosimetric impact of air volume. 3. Results All five NPC patients studied were successfully treated with the IMRT, i.e. disease-free with excellent local control after a median follow-up of 13 months. To compare their PTV dose distributions calculated using the AXB algorithm and the MC method, air cavities inside the PTV were either included or excluded. These air cavities produced higher statistical noise in the MC simulation due to fewer particle interactions in air than in soft tissues. As noted by De Smedt et al. (2007), the exclusion of air cavities in MC simulations resulted in more accurate dose distributions. Of all five patients, Table 1 lists their PTV including air, air inside the PTV, and PTV without air. On average, the air volume accounts for 10.8% of the PTV. To evaluate the accuracy of the TPS algorithm in non-standard 2
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
P.C.Y. Yeh et al.
Table 1 PTV including and excluding air cavities in five NPC patients studied. Case
1 2 3 4 5 Mean
Table 3 Comparisons of mean doses to OAR (Gy) calculated using the AAA, AXB and MC.
PTV
Air in PTV
PTV (no air)
(cm3)
(cm3, %)
(cm3)
295.3 194.8 202.6 155.9 82.0 186.1
12.6 21.9 34.2 12.1 11.5 18.5
282.7 172.9 168.4 143.8 70.5 167.7
(4.3%) (11.2%) (16.9%) (7.8%) (14.0%) (10.8%)
Right cochlea Left cochlea Spine Right optic Left parotid Right parotid Left optic Right eye Left eye
Table 2 Comparisons of PTV dose distributions among results calculated using the AAA, AXB and MC. AAA
AXBm
MCm
AXBw
MCw
PTV70 mean
70.8 Gy (2.2%)
69.9 Gy (0.9%)
69.3 Gy
71.4 Gy (−0.2%)
71.6 Gy
PTV70-air mean
70.9 Gy (3.4%) 97.6% 97.4% 0.10 0.10
70.0 Gy (1.9%) 90.2% 97.1% 0.13 0.13
68.6 Gy
71.7 Gy (0.0%) 96.8% 97.3% 0.13 0.12
71.7 Gy
PTV70 V > 95% PTV70-air V > 95% PTV70 HI PTV70-air HI
90.4% 95.3% 0.14 0.14
AAA
AAA-MCm
AXBm
AXBm-MCm
MCm
58.8 50.8 25.6 19.5 18.6 19.7 17.8 8.3 8.3
3.0 1.1 −4.4 −3.0 −2.7 −1.3 −2.7 −1.0 −0.6
56.4 48.7 27.8 19.1 18.1 19.2 17.3 7.8 7.9
0.6 −1.0 −2.2 −3.4 −3.2 −1.8 −3.2 −1.5 −1.0
55.8 49.7 29.9 22.6 21.3 21.0 20.5 9.3 8.9
the AXB dose deviations in all water-like tissues. For such tissues, it has been proved (Kan et al., 2013a; Siebers et al., 2000) that AAA dose calculations neglecting the material composition produced negligible effects. For bony structures, AXB is more accurate than AAA because AXB computes the absorbed dose considering the material composition.
4. Discussion 97.0% 97.1% 0.10 0.11
De Smedt et al. (2007) concluded that excluding air cavities from the PTV in MC simulations led to a better dose distribution in the IMRT treatment plan. This is because air cavities produce high statistical noise in the MC calculated dose. Their study also noted that lowering the ECUT value in the EGS-based MC code, corresponding to transporting the track-end electrons in straight lines through air instead of locally depositing these electrons in air, did not reduce the noise significantly. In Table 2 it is demonstrated that AAA and AXB resulted in smaller differences compared to MC for the PTV doses including and excluding air cavities. Kan et al. (2013a) assessed the dosimetric impact of AXB using a nine-field sliding window IMRT in NPC. Compared to the AAA data, they observed that the AXB mean dose to PTV70 was 0.9% lower, the HI was 35% higher, and the V > 95% was approximately equal. In our study using a seven-coplanar dynamic-sliding window IMRT in NPC, we found that compared to the AAA data the AXB mean dose to PTV70 was 1.3% (1.4%) lower, the HI was 36% (31%) higher, and the V > 95% was 7.6% (0.3%) lower for the PTV including (excluding) air cavities. We also found that compared to the MC data the AXB mean dose to PTV70 was 0.9% (1.9%) higher, the HI was 1.5% (7.3%) lower, and the V > 95% was 0.2% lower (1.9% higher). These indicate that (1) there are no systematic differences related to the treatment techniques, (2) an overall good agreement was achieved, with differences generally smaller than 2%, for both AAA and AXB in the PTV mean doses, and (3) the AXB was in preference to AAA for the IMRT in NPC. From comparisons of the mean doses to OAR, the AXB resulted in more accurate doses in the bony structures, while AAA produced less underestimated doses in the soft tissues. Although there have been numerous reports comparing dose data among AAA, AXB and MC, most of them dealt with homogeneous and anthropomorphic phantoms using the AXB versions 10 and 11 (Ojala, 2014). The present study applied AXB version 13 in the evaluation of dose calculations under clinical settings.
(IMRT) and non-measurable (patients) conditions, MC simulation is the gold standard for comparisons (Ojala, 2014). Since both AXB and MC calculate electron fluence in the voxel volume and apply the fluence-to-dose function, these comparisons are best made using the dose-to-medium (m) rather than the dose-to-water (w) option. In this way, it allows the detection of differences arising from the dose calculation algorithm but not the medium-to-water dose conversion. Similarly, because AAA applies electron density correction to the water dose kernel, AAA is more similar to calculate the dose-to-medium (Fogliata et al., 2013). Therefore, PTV dose distributions in terms of the dose-to-medium are compared in Table 2 among results calculated using the AAA, AXB and MC. It is observed that AAA overestimates the mean dose to PTV70 by 2.2% or 3.4% in the PTV including or excluding air cavities. Corresponding overestimation by the AXB is 0.9% or 1.9%, respectively. Conversely, ICRU (2010) recommends the use of dose-towater in clinical practices because it better reflects the radiotherapy experience and more closely reflects the cellular composition. Table 2 also compares the PTV dose distributions calculated using AXB and MC in terms of the dose-to-water. A negligible difference of −0.2% or 0.0% in the mean dose to PTV70 is found when the PTV includes or excludes air cavities. From these comparisons, it proves that AXB is more accurate than AAA for dose calculations in the PTV. Excluding air cavities from the PTV resulted in a perfect agreement between AXBw and MCw. Besides the mean doses in PTV, Table 2 also provides data on the HI and V > 95% calculated using AAA, AXB and MC. Again, agreements between AXB and MC data are better than those between AAA and MC data. Further, it shows a better PTV coverage in AXB and MC when air is removed. The values of PTV70-air V > 95% for AXBm, MCm, AXBw and MCw are 97.1%, 95.3%, 97.3% and 97.1%, respectively, larger than the corresponding values of 90.2%, 90.4%, 96.8% and 97.0% for the PTV70 V > 95%. However, AAA shows the opposite trend. Table 3 compares the mean doses to OAR among results calculated using the AAA, AXB and MC, where MC doses are listed in a decreasing order. It is noted that both AAA and AXB underestimate the doses to OAR except cochlea. The dose deviations generally reduce with decreasing MC dose. For cochlea and spine, with a high proportion of bony structures, the AXB dose deviations are less than the AAA dose deviations. On the other hand, the AAA dose deviations are less than
5. Conclusions In the present study, the AXB algorithm was tested for its validity in the IMRT of NPC patients. From comparisons with the MC simulations, it was found that the AXB was in comparable accuracy for dose calculations. It was also found the AXB was more accurate than the AAA in dose calculations within the heterogeneous tissueair-bone region. 3
P.C.Y. Yeh et al.
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
Disclosures
nasopharyngeal carcinomas. Int. J. Radiat. Oncol. Biol. Phys. 85, 73–80. Kan, M.W.K., Yu, P.K.N., Leung, L.H.T., 2013b. A review on the use of grid-based Boltzmann equation solvers for dose calculation in external photon beam treatment planning. Biomed. Res. Int. 2013, 692874. http://dx.doi.org/10.1155/2013/692874. Lin, M.H., Chao, T.C., Lee, C.C., Tung, C.J., Yeh, C.C., Hong, J.H., 2009. Measurement based Monte Carlo dose calculation system for IMRT pretreatment and on-line transit dose verifications. Med. Phys. 36, 1167–1175. Ojala, J., 2014. The accuracy of the Acuros XB algorithm in external beam radiotherapy – a comprehensive review. Int. J. Cancer Ther. Oncol. 2 (4), 020417. http://dx.doi.org/ 10.14319/ijcto.0204.17. Rogers, D.W.O., Faddegon, B.A., Ding, G.X., Ma, C.M., We, J., 1995. BEAM: a Monte Carlo code to simulate radiotherapy treatment units. Med. Phys. 22, 503–524. Siebers, J.V., Keall, P.J., Nahum, A.E., Mohan, R., 2000. Converting absorbed dose to medium to absorbed dose to water for Monte Carlo based photon beam dose calculations. Phys. Med. Biol. 45, 983–995. Sun, X., Su, S., Chen, C., Han, F., Zhao, C., Xiao, W., Deng, X., Huang, S., Lin, C., Lu, T., 2014. Long-term outcomes of intensity- modulated radiotherapy for 868 patients with nasopharyngeal carcinoma: an analysis of survival and treatment toxicities. Radiother. Oncol. 110, 398–403. Tao, C.J., Yi, J.L., Chen, N.Y., Ren, W., Cheng, J., Tung, S., Kong, L., Lin, S.J., Pan, J.J., Zhang, G.S., Hu, J., Qi, Z.Y., Ma, J., Lu, J.D., Yan, D., Sun, Y., 2015. Multi-subject atlas-based auto-segmentation reduces interobserver variation and improves dosimetric parameter consistency for organs at risk in nasopharyngeal carcinoma: A multi-institution clinical study. Radiother. Oncol. 115, 407–411. Torre, L.A., Bray, F., Siegel, R.L., Ferlay, J., Lortet-Tieulent, J., Jemal, A., 2015. Global cancer statistics, 2012. CA Cancer J. Clin. 65, 87–108. http://dx.doi.org/10.3322/ caac.21262. Yeh, C.Y., Lee, C.C., Chao, T.C., Lin, M.H., Lai, P.A., Liu, F.H., Tung, C.J., 2014. Application of the measurement-based Monte Carlo method in nasopharyngeal cancer patients for intensity modulated radiation therapy. Radiat. Phys. Chem. 95, 240–242.
The authors report no conflicts of interest. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References Bush, K., Gagne, I.M., Zavgorodni, S., Ansbacher, W., Beckham, W., 2011. Dosimetric validation of Acuros XB with Monte Carlo methods for photon dose calculations. Med. Phys. 38, 2208–2221. De Smedt, B., Vanderstraeten, B., Reynaert, N., De Gersem, W., De Neve, W., Thierens, H., 2007. The influence of air cavities within the PTV on Monte Carlo based IMRT optimization. J. Phys.: Conf. Ser. 74 (021003), 1–8. Failla, G.A., Wareing, T., Archambault, Y., Thompson, S., 2010. Acuros XB advanced dose calculation for the Eclipse treatment planning system. Palo Alto (CA), Varian Medical Systems Report No. RAD 10156. 〈https://www.varian.com/sites/default/ files/resource_attachments/AcurosXBClinicalPerspectives_0.pdf〉. Fogliata, A., Scorsetti, M., Navarria, P., Catalano, M., Clivio, A., Cozzi, L., Lobefalo, F., Nicolini, G., Palumbo, V., Pellegrini, C., Reggiori, G., Roggio, A., Vanetti, E., Alongi, F., Pentimalli, S., Mancosu, P., 2013. Dosimetric comparison between VMAT with different dose calculation algorithms and protons for soft-tissue sarcoma radiotherapy. Acta Oncol. 52, 545–552. Han, T., Mikell, J.K., Salehpour, M., Mourtada, F., 2011. Dosimetric comparison of Acuros XB deterministic radiation transport method with Monte Carlo and modelbased convolution methods in heterogeneous media. Med. Phys. 38, 2651–2664. ICRU, 2010. Prescribing recording and reporting photon beam intensity modulated radiation therapy (IMRT). ICRU Report 83. Oxford, Oxford University Press. Kan, M.W.K., Leung, L.H.T., Yu, P.K.N., 2013a. Dosimetric impact of using the Acuros XB algorithm for intensity modulated radiation therapy and RapidArc planning in
4
Contents lists available at ScienceDirect
Radiation Physics and Chemistry journal homepage: www.elsevier.com/locate/radphyschem
Monte Carlo evaluation of Acuros XB dose calculation Algorithm for intensity modulated radiation therapy of nasopharyngeal carcinoma ⁎
Peter C.Y. Yeha,b, C.C. Leec,d, T.C. Chaoc,e, C.J. Tungc,e, a
Department of Radiation Oncology, Tungs MetroHarbor General Hospital, Taichung, Taiwan Jen-Teh Junior College of Medicine, Nursing and Management, Miaoli, Taiwan c Department of Medical Imaging and Radiological Sciences, College of Medicine, Chang Gung University, Taiwan d Department of Radiation Oncology, Chang Gung Memorial Hospital, Taiwan e Institute for Radiological Research, Chang Gung University/Chang Gung Memorial Hospital, Taiwan b
A R T I C L E I N F O
A BS T RAC T
Keywords: Acuros XB Monte Carlo Anisotropic analytical algorithm Nasopharyngeal carcinoma Intensity modulated radiation therapy
Intensity-modulated radiation therapy is an effective treatment modality for the nasopharyngeal carcinoma. One important aspect of this cancer treatment is the need to have an accurate dose algorithm dealing with the complex air/bone/tissue interface in the head-neck region to achieve the cure without radiation-induced toxicities. The Acuros XB algorithm explicitly solves the linear Boltzmann transport equation in voxelized volumes to account for the tissue heterogeneities such as lungs, bone, air, and soft tissues in the treatment field receiving radiotherapy. With the single beam setup in phantoms, this algorithm has already been demonstrated to achieve the comparable accuracy with Monte Carlo simulations. In the present study, five nasopharyngeal carcinoma patients treated with the intensity-modulated radiation therapy were examined for their dose distributions calculated using the Acuros XB in the planning target volume and the organ-at-risk. Corresponding results of Monte Carlo simulations were computed from the electronic portal image data and the BEAMnrc/DOSXYZnrc code. Analysis of dose distributions in terms of the clinical indices indicated that the Acuros XB was in comparable accuracy with Monte Carlo simulations and better than the anisotropic analytical algorithm for dose calculations in real patients.
1. Introduction Nasopharyngeal carcinoma (NPC) is a unique primary head and neck cancer arising from the nasopharynx with an estimated 86,700 new cases and 50,800 deaths in 2012 worldwide. High incidence rates are observed in southeastern China, Hong Kong, Malaysia, Indonesia and Singapore (Torre et al., 2015). Intensity-modulated radiation therapy (IMRT) with concurrent CDDP+5-FU chemotherapy is the primary treatment of choice for NPC because of its unique anatomical location and sensitivity to radiotherapy (Sun et al., 2014; Tao et al., 2015). IMRT provides the steep radiation dose gradient to produce a high degree of conformal tumor target coverage and a sparing of normal tissues. The accurate dose calculations are required to obtain therapeutic advantages of the IMRT. One important aspect of IMRT for NPC patients is the need to have an accurate dose calculation algorithm to deal with the effects on the complex air/bone/tissue interface for achieving the cure without radiation-induced toxicities. Although Monte Carlo (MC) calculations had the best agreement with measured data within the inhomogeneous
⁎
region (Ojala, 2014), these calculations require a large number of individual particles transporting in matter, thus resulting in an extensive computing time. This makes MC simulations virtually impractical in the clinical environment even with the variance reduction method, efficient sampling and coding technique (Bush et al., 2011). The anisotropic analytical algorithm (AAA), a convolution/ superposition method, has been widely utilized for dose calculations in the treatment planning system (TPS). However, AAA was reported to lead to dose discrepancies in the air/bone/tissue region due to its inherent pencil beam kernel and independent depth/lateral scaling of the kernel for heterogeneity correction (Kan et al., 2013a). The Acuros XB (AXB) algorithm provides a new advanced dose calculation method (Varian Medical Systems, Palo Alto, CA) applied in the TPS. It explicitly solves the linear Boltzmann transport equation by a deterministic method using discretized cross sections as radiations interact with the voxel volumes in matter. AXB makes use of the chemical composition of each material in the volume during radiation transport (Kan et al., 2013a, 2013b). Therefore, it directly accounts for the effects on tissue heterogeneities. With the single beam setup in
Correspondence to: Department of Medical Imaging and Radiological Sciences, Chang Gung University, 259 Wenhua 1st Rd., Guishan Dist., Taoyuan City 33302, Taiwan. E-mail addresses: [email protected], [email protected] (C.J. Tung).
http://dx.doi.org/10.1016/j.radphyschem.2017.02.025 Received 17 September 2016; Received in revised form 8 February 2017; Accepted 10 February 2017 0969-806X/ © 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Yeh, P.C.Y., Radiation Physics and Chemistry (2017), http://dx.doi.org/10.1016/j.radphyschem.2017.02.025
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
P.C.Y. Yeh et al.
phantoms, AXB has already been demonstrated to achieve the required accuracy compared with MC simulations but with faster computing speed and without statistical noise (Failla et al., 2010). For instance, Bush et al. (2011) compared AXB and MC calculated doses in a heterogeneous air/bone/lung phantom for 6 and 18 MV photon beams. They showed the maximum discrepancies were within 2% in lung, 1.8% in bone and 4.5% in air. Han et al. (2011) noted that the agreement between AXB and MC doses in a soft tissue/bone/lung phantom for 6 MV photon beam was within 2%. Although the accuracy of AXB has been shown in heterogeneous phantoms, the clinical validation of AXB in real patients treated with IMRT should be established, especially for tumors in the head and neck region where a complex air/bone/soft tissue interface could cause dose perturbations. The present study aims to compare dose distributions calculated using the AXB algorithm and the MC simulation and to evaluate the clinical impact of AXB on NPC patients treated with the IMRT. Five NPC patients were studied with measurement-based MC (MBMC) calculations applying the electronic portal image data and the BEAMnrc/DOSXYZnrc code. Analysis of dose distributions in the planning target volume (PTV) and the organs-at-risk (OAR) were made. It indicated that the AXB algorithm was in comparable accuracy with the Monte Carlo simulation in the head and neck region. 2. Materials and methods The AXB algorithm solves the linear Boltzmann transport equation of coupled photon and electron fluences in a voxel volume of matter through several steps. First, the external photon and electron sources are transported into the volume using ray-tracing techniques. Then, a finite-element method is used to find the energy- and angulardependent fluences in this volume by applying discretized cross sections and stopping powers. Since the AXB utilizes mass densities and atomic compositions in the interaction cross sections, its calculation is analogous to the MC simulation. Comparing to the stochastic MC simulation, however, the AXB greatly reduces the computing time because of its deterministic computation. In the present study, the AXB version 13.026 in the Eclipse TPS (Varian Medical Systems, Palo Alto, CA) was used with a grid size of 2 mm2. The AXB modeled four radiation sources: (1) primary source or bremsstrahlung photons created in the target, (2) extra focal source or photons resulting from interactions in the accelerator head (the flattening filter, primary collimators, and secondary jaws), (3) electron contamination, and (4) photons scattered from wedges (Failla et al., 2010). Five NPC patients treated with the IMRT received cumulative doses of 70 Gy in 35 fractions (Yeh et al., 2014). Seven coplanar 6 MV photon beams were used with the dynamic sliding window technique through a Millennium 120-leaf multi-leaf collimator (MLC) in the Varian 21EX linear accelerator. MC simulations were performed using the MBMC method, which applied the electronic portal image data and the BEAMnrc/DOSXYZnrc code version 2007 (Rogers et al., 1995). The MBMC method has been described in detail previously (Lin et al., 2009) and is summarized here with the aid of Fig. 1. First, an incident parallel circular electron beam with a Gaussian intensity distribution of full-width at half-maximum equal to 0.12 cm and a mean electron energy of 6.3 MeV with 3% energy spread was incident onto the Varian 21EX linear accelerator tungsten target to generate an open-field phase-space file, i.e. data on energies, positions, directions, and weightings of every particles crossing the scoring plane at 80 cm from the source. Then, MC simulations were performed by applying the variance reduction technique with the directional bremsstrahlung splitting. The splitting-field source-to-surface distance was set at 100 cm. The splitting-field radius was equal to the field size. The Russian roulette plane was chosen above the bottom of the flattening filter. Here at least 3.0×107 particles were simulated in each IMRT field to reduce the uncertainty to ≤2% in the phase-space file. Next, an efficiency map of each IMRT field was obtained from data
Fig. 1. The MBMC simulation setup.
collected on the amorphous silicon aS1000 electronic portal imaging device to adjust the weighting of each particle in the phase-space file. Further, the dose distribution in the patient was calculated using the DOSXYZ code (Rogers et al., 1995) and the efficiency map of IMRT photon beams transported through the patient. Finally, the calculated dose-to-medium could be converted to dose-to-water using the waterto-medium stopping power ratios (Siebers et al., 2000). To allow uncertainties in the patient positioning, alignment and respiratory motion during the IMRT, PTV was determined to be the irradiated tumor volume plus a 3–5 mm margin. The PTV dose distribution was evaluated by V > 95%, the percent PTV volume receiving ≥95% of the prescribed dose (70 Gy). The homogeneity index (HI) was evaluated by the ratio (D2%-D98%)/D50%, where D2%, D50% and D98% are the minimum doses received by 2%, 50% and 98% of the PTV volume, respectively. A lower HI indicated a better dose homogeneity or less cold/hot spots (Kan et al., 2013a). Although air cavities were usually included in the PTV, a common practice among radiation oncologists during the treatment planning, PTV, V > 95% and HI were all determined by either including or excluding air cavities in the target volume in order to assess the dosimetric impact of air volume. 3. Results All five NPC patients studied were successfully treated with the IMRT, i.e. disease-free with excellent local control after a median follow-up of 13 months. To compare their PTV dose distributions calculated using the AXB algorithm and the MC method, air cavities inside the PTV were either included or excluded. These air cavities produced higher statistical noise in the MC simulation due to fewer particle interactions in air than in soft tissues. As noted by De Smedt et al. (2007), the exclusion of air cavities in MC simulations resulted in more accurate dose distributions. Of all five patients, Table 1 lists their PTV including air, air inside the PTV, and PTV without air. On average, the air volume accounts for 10.8% of the PTV. To evaluate the accuracy of the TPS algorithm in non-standard 2
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
P.C.Y. Yeh et al.
Table 1 PTV including and excluding air cavities in five NPC patients studied. Case
1 2 3 4 5 Mean
Table 3 Comparisons of mean doses to OAR (Gy) calculated using the AAA, AXB and MC.
PTV
Air in PTV
PTV (no air)
(cm3)
(cm3, %)
(cm3)
295.3 194.8 202.6 155.9 82.0 186.1
12.6 21.9 34.2 12.1 11.5 18.5
282.7 172.9 168.4 143.8 70.5 167.7
(4.3%) (11.2%) (16.9%) (7.8%) (14.0%) (10.8%)
Right cochlea Left cochlea Spine Right optic Left parotid Right parotid Left optic Right eye Left eye
Table 2 Comparisons of PTV dose distributions among results calculated using the AAA, AXB and MC. AAA
AXBm
MCm
AXBw
MCw
PTV70 mean
70.8 Gy (2.2%)
69.9 Gy (0.9%)
69.3 Gy
71.4 Gy (−0.2%)
71.6 Gy
PTV70-air mean
70.9 Gy (3.4%) 97.6% 97.4% 0.10 0.10
70.0 Gy (1.9%) 90.2% 97.1% 0.13 0.13
68.6 Gy
71.7 Gy (0.0%) 96.8% 97.3% 0.13 0.12
71.7 Gy
PTV70 V > 95% PTV70-air V > 95% PTV70 HI PTV70-air HI
90.4% 95.3% 0.14 0.14
AAA
AAA-MCm
AXBm
AXBm-MCm
MCm
58.8 50.8 25.6 19.5 18.6 19.7 17.8 8.3 8.3
3.0 1.1 −4.4 −3.0 −2.7 −1.3 −2.7 −1.0 −0.6
56.4 48.7 27.8 19.1 18.1 19.2 17.3 7.8 7.9
0.6 −1.0 −2.2 −3.4 −3.2 −1.8 −3.2 −1.5 −1.0
55.8 49.7 29.9 22.6 21.3 21.0 20.5 9.3 8.9
the AXB dose deviations in all water-like tissues. For such tissues, it has been proved (Kan et al., 2013a; Siebers et al., 2000) that AAA dose calculations neglecting the material composition produced negligible effects. For bony structures, AXB is more accurate than AAA because AXB computes the absorbed dose considering the material composition.
4. Discussion 97.0% 97.1% 0.10 0.11
De Smedt et al. (2007) concluded that excluding air cavities from the PTV in MC simulations led to a better dose distribution in the IMRT treatment plan. This is because air cavities produce high statistical noise in the MC calculated dose. Their study also noted that lowering the ECUT value in the EGS-based MC code, corresponding to transporting the track-end electrons in straight lines through air instead of locally depositing these electrons in air, did not reduce the noise significantly. In Table 2 it is demonstrated that AAA and AXB resulted in smaller differences compared to MC for the PTV doses including and excluding air cavities. Kan et al. (2013a) assessed the dosimetric impact of AXB using a nine-field sliding window IMRT in NPC. Compared to the AAA data, they observed that the AXB mean dose to PTV70 was 0.9% lower, the HI was 35% higher, and the V > 95% was approximately equal. In our study using a seven-coplanar dynamic-sliding window IMRT in NPC, we found that compared to the AAA data the AXB mean dose to PTV70 was 1.3% (1.4%) lower, the HI was 36% (31%) higher, and the V > 95% was 7.6% (0.3%) lower for the PTV including (excluding) air cavities. We also found that compared to the MC data the AXB mean dose to PTV70 was 0.9% (1.9%) higher, the HI was 1.5% (7.3%) lower, and the V > 95% was 0.2% lower (1.9% higher). These indicate that (1) there are no systematic differences related to the treatment techniques, (2) an overall good agreement was achieved, with differences generally smaller than 2%, for both AAA and AXB in the PTV mean doses, and (3) the AXB was in preference to AAA for the IMRT in NPC. From comparisons of the mean doses to OAR, the AXB resulted in more accurate doses in the bony structures, while AAA produced less underestimated doses in the soft tissues. Although there have been numerous reports comparing dose data among AAA, AXB and MC, most of them dealt with homogeneous and anthropomorphic phantoms using the AXB versions 10 and 11 (Ojala, 2014). The present study applied AXB version 13 in the evaluation of dose calculations under clinical settings.
(IMRT) and non-measurable (patients) conditions, MC simulation is the gold standard for comparisons (Ojala, 2014). Since both AXB and MC calculate electron fluence in the voxel volume and apply the fluence-to-dose function, these comparisons are best made using the dose-to-medium (m) rather than the dose-to-water (w) option. In this way, it allows the detection of differences arising from the dose calculation algorithm but not the medium-to-water dose conversion. Similarly, because AAA applies electron density correction to the water dose kernel, AAA is more similar to calculate the dose-to-medium (Fogliata et al., 2013). Therefore, PTV dose distributions in terms of the dose-to-medium are compared in Table 2 among results calculated using the AAA, AXB and MC. It is observed that AAA overestimates the mean dose to PTV70 by 2.2% or 3.4% in the PTV including or excluding air cavities. Corresponding overestimation by the AXB is 0.9% or 1.9%, respectively. Conversely, ICRU (2010) recommends the use of dose-towater in clinical practices because it better reflects the radiotherapy experience and more closely reflects the cellular composition. Table 2 also compares the PTV dose distributions calculated using AXB and MC in terms of the dose-to-water. A negligible difference of −0.2% or 0.0% in the mean dose to PTV70 is found when the PTV includes or excludes air cavities. From these comparisons, it proves that AXB is more accurate than AAA for dose calculations in the PTV. Excluding air cavities from the PTV resulted in a perfect agreement between AXBw and MCw. Besides the mean doses in PTV, Table 2 also provides data on the HI and V > 95% calculated using AAA, AXB and MC. Again, agreements between AXB and MC data are better than those between AAA and MC data. Further, it shows a better PTV coverage in AXB and MC when air is removed. The values of PTV70-air V > 95% for AXBm, MCm, AXBw and MCw are 97.1%, 95.3%, 97.3% and 97.1%, respectively, larger than the corresponding values of 90.2%, 90.4%, 96.8% and 97.0% for the PTV70 V > 95%. However, AAA shows the opposite trend. Table 3 compares the mean doses to OAR among results calculated using the AAA, AXB and MC, where MC doses are listed in a decreasing order. It is noted that both AAA and AXB underestimate the doses to OAR except cochlea. The dose deviations generally reduce with decreasing MC dose. For cochlea and spine, with a high proportion of bony structures, the AXB dose deviations are less than the AAA dose deviations. On the other hand, the AAA dose deviations are less than
5. Conclusions In the present study, the AXB algorithm was tested for its validity in the IMRT of NPC patients. From comparisons with the MC simulations, it was found that the AXB was in comparable accuracy for dose calculations. It was also found the AXB was more accurate than the AAA in dose calculations within the heterogeneous tissueair-bone region. 3
P.C.Y. Yeh et al.
Radiation Physics and Chemistry (xxxx) xxxx–xxxx
Disclosures
nasopharyngeal carcinomas. Int. J. Radiat. Oncol. Biol. Phys. 85, 73–80. Kan, M.W.K., Yu, P.K.N., Leung, L.H.T., 2013b. A review on the use of grid-based Boltzmann equation solvers for dose calculation in external photon beam treatment planning. Biomed. Res. Int. 2013, 692874. http://dx.doi.org/10.1155/2013/692874. Lin, M.H., Chao, T.C., Lee, C.C., Tung, C.J., Yeh, C.C., Hong, J.H., 2009. Measurement based Monte Carlo dose calculation system for IMRT pretreatment and on-line transit dose verifications. Med. Phys. 36, 1167–1175. Ojala, J., 2014. The accuracy of the Acuros XB algorithm in external beam radiotherapy – a comprehensive review. Int. J. Cancer Ther. Oncol. 2 (4), 020417. http://dx.doi.org/ 10.14319/ijcto.0204.17. Rogers, D.W.O., Faddegon, B.A., Ding, G.X., Ma, C.M., We, J., 1995. BEAM: a Monte Carlo code to simulate radiotherapy treatment units. Med. Phys. 22, 503–524. Siebers, J.V., Keall, P.J., Nahum, A.E., Mohan, R., 2000. Converting absorbed dose to medium to absorbed dose to water for Monte Carlo based photon beam dose calculations. Phys. Med. Biol. 45, 983–995. Sun, X., Su, S., Chen, C., Han, F., Zhao, C., Xiao, W., Deng, X., Huang, S., Lin, C., Lu, T., 2014. Long-term outcomes of intensity- modulated radiotherapy for 868 patients with nasopharyngeal carcinoma: an analysis of survival and treatment toxicities. Radiother. Oncol. 110, 398–403. Tao, C.J., Yi, J.L., Chen, N.Y., Ren, W., Cheng, J., Tung, S., Kong, L., Lin, S.J., Pan, J.J., Zhang, G.S., Hu, J., Qi, Z.Y., Ma, J., Lu, J.D., Yan, D., Sun, Y., 2015. Multi-subject atlas-based auto-segmentation reduces interobserver variation and improves dosimetric parameter consistency for organs at risk in nasopharyngeal carcinoma: A multi-institution clinical study. Radiother. Oncol. 115, 407–411. Torre, L.A., Bray, F., Siegel, R.L., Ferlay, J., Lortet-Tieulent, J., Jemal, A., 2015. Global cancer statistics, 2012. CA Cancer J. Clin. 65, 87–108. http://dx.doi.org/10.3322/ caac.21262. Yeh, C.Y., Lee, C.C., Chao, T.C., Lin, M.H., Lai, P.A., Liu, F.H., Tung, C.J., 2014. Application of the measurement-based Monte Carlo method in nasopharyngeal cancer patients for intensity modulated radiation therapy. Radiat. Phys. Chem. 95, 240–242.
The authors report no conflicts of interest. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References Bush, K., Gagne, I.M., Zavgorodni, S., Ansbacher, W., Beckham, W., 2011. Dosimetric validation of Acuros XB with Monte Carlo methods for photon dose calculations. Med. Phys. 38, 2208–2221. De Smedt, B., Vanderstraeten, B., Reynaert, N., De Gersem, W., De Neve, W., Thierens, H., 2007. The influence of air cavities within the PTV on Monte Carlo based IMRT optimization. J. Phys.: Conf. Ser. 74 (021003), 1–8. Failla, G.A., Wareing, T., Archambault, Y., Thompson, S., 2010. Acuros XB advanced dose calculation for the Eclipse treatment planning system. Palo Alto (CA), Varian Medical Systems Report No. RAD 10156. 〈https://www.varian.com/sites/default/ files/resource_attachments/AcurosXBClinicalPerspectives_0.pdf〉. Fogliata, A., Scorsetti, M., Navarria, P., Catalano, M., Clivio, A., Cozzi, L., Lobefalo, F., Nicolini, G., Palumbo, V., Pellegrini, C., Reggiori, G., Roggio, A., Vanetti, E., Alongi, F., Pentimalli, S., Mancosu, P., 2013. Dosimetric comparison between VMAT with different dose calculation algorithms and protons for soft-tissue sarcoma radiotherapy. Acta Oncol. 52, 545–552. Han, T., Mikell, J.K., Salehpour, M., Mourtada, F., 2011. Dosimetric comparison of Acuros XB deterministic radiation transport method with Monte Carlo and modelbased convolution methods in heterogeneous media. Med. Phys. 38, 2651–2664. ICRU, 2010. Prescribing recording and reporting photon beam intensity modulated radiation therapy (IMRT). ICRU Report 83. Oxford, Oxford University Press. Kan, M.W.K., Leung, L.H.T., Yu, P.K.N., 2013a. Dosimetric impact of using the Acuros XB algorithm for intensity modulated radiation therapy and RapidArc planning in
4