- Email: [email protected]
The Accuracy of Inhomogeneity Corrections in Intensity Modulated Radiation Therapy Planning in Philips Pinnacle System
Medical Dosimetry, Vol. 36, No. 3, pp. 240-245, 2011 Copyright © 2011 American Association of Medical Dosimetrists Printed in the USA. All rights reserved 0958-3947/11/$–see front matter
doi:10.1016/j.meddos.2010.03.010
THE ACCURACY OF INHOMOGENEITY CORRECTIONS IN INTENSITY MODULATED RADIATION THERAPY PLANNING IN PHILIPS PINNACLE SYSTEM PARHAM ALAEI, PH.D., and PATRICK D. HIGGINS, PH.D. Department of Therapeutic Radiology-Radiation Oncology, University of Minnesota, Minneapolis, MN (Received 3 February 2010; accepted 29 March 2010)
Abstract—The degree of accuracy of inhomogeneity corrections in a treatment planning system is dependent on the algorithm used by the system. The choice of field size, however, could have an effect on the calculation accuracy as well. There have been several evaluation studies on the accuracy of inhomogeneity corrections used by different algorithms. Most of these studies, however, focus on evaluating the dose in phantom using simplified geometry and open/static fields. This work focuses on evaluating the degree of dose accuracy in calculations involving intensity-modulated radiation therapy (IMRT) fields incident on a phantom containing both lung- and bone-equivalent heterogeneities using 6 and 10 MV beams. IMRT treatment plans were generated using the Philips Pinnacle treatment planning system and delivered to a phantom containing 55 thermoluminescent dosimeter (TLD) locations within the lung and bone and near the lung and bone interfaces with solid water. The TLD readings were compared with the dose predicted by the planning system. We find satisfactory agreement between planned and delivered doses, with an overall absolute average difference between measurement and calculation of 1.2% for the 6 MV and 3.1% for the 10 MV beam with larger variations observed near the interfaces and in areas of high-dose gradient. The results presented here demonstrate that the convolution algorithm used in the Pinnacle treatment planning system produces accurate results in IMRT plans calculated and delivered to inhomogeneous media, even in regions that potentially lack electronic equilibrium. © 2011 American Association of Medical Dosimetrists. Key Words: Heterogeneity corrections, IMRT, Thermoluminescent dosimetry.
INTRODUCTION
Intensity-modulated radiation therapy (IMRT) treatment planning, because of the presence of small fields, introduces new challenges in dose calculation accuracy, both in homogeneous and inhomogeneous media. Conventional treatment planning rarely involves fields smaller than 4 ⫻ 4 cm2, with the exception of radiosurgery, whereas IMRT planning often involves fields smaller than 4 ⫻ 4 cm2, possibly as small as 1 ⫻ 1 cm2, depending on the system settings and/or preferences set by the user. This can lead to conditions of electronic disequilibrium, which may not be accurately accounted for by the treatment planning system.23 The degree to which a treatment planning system can predict the dose accurately for small fields has been evaluated by a few investigators.5,7,10,17,18,24 We have evaluated the accuracy of the Philips Pinnacle3 treatment planning system (Philips Medical Systems, Milpitas, CA) in modeling the dose within and near the interface of inhomogeneities for IMRT plans in phantom. The collapsed cone convolution algorithm employed in Pinnacle25 is known to have a higher degree of accuracy in inhomogeneous media than other algorithms.5,7,23 Ahnesjo26 has shown that collapsed cone convolution can calculate the dose in inhomogeneous media with very high accuracy, provided charged particle equilibrium is present. So, although collapsed cone convolution is one of the most accurate methods for dose
Inhomogeneity corrections are commonly used in computerized treatment planning. Depending on the algorithm, there are varying degrees to which a treatment planning system can accurately predict the dose in and around an inhomogeneity. There are many publications on the evaluation of inhomogeneity corrections in treatment planning systems.1–18 Most of these publications evaluate this in simple geometry, e.g., in slab phantoms with the addition of lung- or bone-equivalent slabs in between polystyrene or solid water slabs.2–5,9,10,14,16 –18 There have also been a few investigations evaluating the dose in humanoid phantoms12,15,16 and a few patient studies.6,8,11,13 The 2 common methods of evaluation compare the planpredicted dose with measurements5,12,15–19 and with Monte Carlo calculations.6 –11,13 Among the different treatment planning algorithms, the ones using pencil-beam algorithms with correctionbased inhomogeneity calculations are known to suffer from varying degrees of inaccuracies in predicting the dose in and around inhomogeneities.7,20 –22 Systems using convolution-based algorithms or Monte Carlo are known to predict the dose more accurately.7,21,23 Reprint requests to: Parham Alaei, Ph.D., Department of Therapeutic Radiology-Radiation Oncology, University of Minnesota, 420 Delaware Street SE, MMC 494, Minneapolis, MN 55455. E-mail: [email protected] 240
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
AND
P.D. HIGGINS
241
finding better accuracy at the lung–soft tissue interfaces than at the bone–soft tissue ones for field sizes as small as 2 ⫻ 2 cm2. This study compares Pinnacle-predicted doses to measurements in a phantom containing both lung- and boneequivalent heterogeneities for intensity-modulated beams. MATERIALS AND METHODS
Fig. 1. The phantom used for the study with the lung and bone inserts shown in the insets. Locations of initial TLD slots and additional milled ones are indicated in the figure. The overall phantom size is 30 cm high by 35 cm wide by 13 cm thick. The lung-equivalent insert has a square cross-section of 6.5 ⫻ 6.5 cm and the bone-equivalent one is a cylinder of diameter 3 cm.
calculation in inhomogeneous materials, this limitation warrants an investigation into the degree of accuracy of inhomogeneity corrections in the Pinnacle system. This can only be evaluated with detailed measurements. Previously, Martens et al.,7 Jones and Das,10 AlHallaq et al.,12 and Davidson et al.15,19 have done comparison studies that included the Pinnacle system. Martens et al.7 showed that although collapsed cone convolution does not predict the dose accurately at the air-phantom interface, it is an improvement over pencil beam algorithms. Jones and Das10 compared the dose in a phantom with low-density heterogeneities predicted by various algorithms, including collapsed cone convolution, with that calculated using Monte Carlo and concluded that collapsed cone convolution models the dose more accurately in lung than other methods evaluated. Al-Hallaq et al.12 compared the dose predicted by the CORVUS planning system (Best NOMOS, Pittsburgh, PA) with thermoluminescent dosimeters (TLDs) placed in the Rando head phantom. They then imported the multileaf collimator (MLC) leaf sequences into Pinnacle and calculated the dose. However, the earlier version of Pinnacle used in that study does not account for MLC leakage and rounded leaf edge. They reported good agreement between measured and Pinnacle-predicted doses. Davidson et al.15 evaluated the accuracy of dose calculations by CORVUS and Pinnacle systems in a thorax phantom and concluded that Pinnacle predicts the dose to within 5% of measured dose. In another paper, Davidson et al.19 evaluated the accuracy of 5 treatment planning systems, including Pinnacle, in an anthropomorphic thorax phantom. Saxena and Higgins24 compared measured and Pinnacle-computed doses in a phantom including lung and bone inhomogeneities for a range of square field sizes,
Phantom The phantom used for evaluation is a modified Benchmark IMRT QA phantom made by Civco (Civco Medical Solutions, Kalona, IA). The phantom includes lung- and bone-equivalent inhomogeneities, with a number of slots for TLD placement in the lung-equivalent insert. To measure the dose within and near these inhomogeneities, additional slots of 4.5-mm diameter and 1-mm depth were milled in the lung-equivalent insert as well as in and near the bone-equivalent insert (Fig. 1). The slots’ dimensions allow for placement of lithium fluoride (LiF) TLD chips, with a minimal amount of air between the chips and the phantom. Treatment planning Pinnacle treatment planning system v8.0m was used to generate IMRT plans on the phantom. The planning system was commissioned for the 6 and 10 MV beams, and modeling was performed, resulting in excellent agreement between modeled and measured profiles for both energies. The agreement between modeled and measured profiles is better than 1% in points beyond buildup region for depth dose curves and within the field for cross profiles. The accuracy of the dose calculation algorithm was tested using simple geometry according to the test described by Tang et al.2 The measurements
Fig. 2. The transverse CT slice of the phantom indicating locations of the target and avoidance structures and the inhomogeneities.
242
Medical Dosimetry
Volume 36, Number 3, 2011
The optimization was done using the Direct Machine Parameter Optimization (DMPO) module of the Pinnacle system and the parameters used were: maximum number of segments—70; minimum segment area—2 cm2; and minimum segment monitor units—5. All optimization and calculations were done heterogeneously with a dose grid of 3 mm3. The calculations were also repeated using a dose grid of 2 mm3 for comparison. A routine IMRT quality assurance (QA), consisting of point dose measurements and film dosimetry, was then performed on the plans. To obtain the expected dose at measurement locations, each TLD was contoured on the corresponding CT slice and the mean and standard deviation of the dose at each TLD position was obtained from the planning system. Fig. 3. The 6 MV IMRT dose distribution within phantom at the CT slice corresponding to the TLD locations within lungequivalent inhomogeneity. The isodose lines are at 10-cGy increments.
were within 1.5% of those reported in this paper for 6 MV and within 2.4% for 10 MV, the larger discrepancies being in the build-up region and at the interfaces. The planning system was also evaluated as part of the work by Saxena and Higgins.24 The phantom was computed tomography (CT)– scanned and 2 7-field IMRT plans were generated by optimizing to a spherical target within the phantom, between the lung and bone inserts. Three additional avoidance structures were added to achieve a realistic IMRT plan with significant modulation (Fig. 2). The inhomogeneities were not used as constraints in the optimization. One plan used 6- and the other used 10 MV beams from a Varian 21ex linear accelerator (Varian Medical Systems, Palo Alto, CA) equipped with 60-pair millennium MLCs.
Dosimeters LiF TLD chips of dimensions 3.2 ⫻ 3.2 ⫻ 0.9 mm were used for dose measurements. A batch of TLDs was calibrated by exposing them to a known dose 3 times, averaging the readings, and calculating the percent difference between each reading and the average of 3 readings. Nitrogen gas flow was used to stabilize the readings. After each irradiation, the chips were left for 24 hours before reading using a Harshaw Model 3500 TLD reader (Thermo Scientific, Waltham, MA) and were post-annealed at 400 °C for 1 hour. The chips with the least amount of variation of response between the 3 measurements (standard deviation of ⬍3%) were chosen and an average calibration factor (cGy/nC) was calculated for each chip. The chips were calibrated for both beam energies and the calibrations were verified throughout the experiment. Measurements The planned IMRT was delivered to phantom, with the calibrated TLD chips placed in a total of 55 locations,
Fig. 4. The dose distribution of the 6 MV plan in phantom within and near the inhomogeneities. Locations of TLD measurement points are indicated, along with the point labels. For the points inside lung-equivalent inhomogeneity (a), there are 6 rows of 5 TLD locations each (points A1–E6) and an additional 2 columns of 4 TLDs each on each side (points a1–a4 and b1– b4). For the points inside and near the bone-equivalent inhomogeneity (b), there is 1 point at the center (0); another 8 points inside the inhomogeneity labeled 1– 8, with number 1 on the top and sequencing clockwise; and another 8 points outside the inhomogeneity labeled 9 –16, with number 9 on top and sequencing clockwise as well.
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
AND
P.D. HIGGINS
243
Fig. 5. The dose distribution for the 10 MV plan in phantom within and near inhomogeneities. Locations of TLD measurements are described in the Fig. 4 legend.
38 inside the lung insert, 18 near the lung–solid water interface, and the remaining 20 inside the lung. Seventeen TLDs were placed in the bone insert and in the surrounding solid water near the bone–solid water interface. All TLDs were placed inside the milled slots in direct contact with phantom material. The phantom was positioned for treatment with the aid of external radioopaque markers placed on it during CT scan. The measurements were performed at least 7 times for each energy/plan. Linear accelerator output during the repeat irradiations was within 0.5% and not accounted for in these measurements. Separate TLD measurements were also performed inside the solid water section of the phantom (upper right of Fig. 1) to evaluate measurement uncertainty far from the interfaces. RESULTS Figure 3 shows the dose distributions within phantom for the 6 MV plan. The contours showing the locations of the TLDs in lung can be seen in the figures. The isodose lines are at 10-cGy increments indicating areas of small- and large-dose gradient. The dose distribution for the 10 MV plan is similar. Figures 4 and 5 are enlarged versions of the 6 and 10 MV plans, respecTable 1. The results of comparison of measured to calculated doses for the IMRT plan using 6 MV beam
All points All lung points Points inside lung Points inside lung near the solid water interface All bone points Points inside bone and near the solid water interface Points inside solid water near the bone interface
Average Percentage Difference
Standard Deviation
⫺1.16 ⫺0.16 ⫺0.14 ⫺0.17
4.98 4.18 2.00 5.91
⫺3.39 0.05
5.96 3.03
⫺7.81
5.23
A positive value indicates the measurement is greater than calculation.
tively, showing the isodose distribution within and near the inhomogeneities and locations of the TLD measurement points. A summary of the TLD measurements for the 6 MV plan can be found in Table 1. The table shows the percentage difference between measured and planned doses at each location. Listed measured doses represent the average of 7 measurements, with corresponding standard deviations. Calculations are the mean dose within each contoured TLD position along with standard deviations as reported by the planning system. The lung measurement points have been divided into points inside the lung (⬎1.5 cm from lung–solid water interface) and points near the solid water interface (⬍1.0 cm from the lung–solid water interface). The bone measurements have been divided into points inside bone and near (⬃0.5 cm) the solid water interface and those in solid water near (⬃0.5 cm) the bone interface. Only one point in the center of the bone insert can be considered far (⬃1.5 cm) from the interface. We find there to be generally good agreement between measured and calculated values within the lung (– 0.14 ⫾ 2.00%), with an increased uncertainty in points near the lung–solid water interface (– 0.17 ⫾ 5.91%), some of which also fall within high-dose gradient areas Table 2. The results of comparison of measured to calculated doses for the IMRT plan using 10 MV beam Average Percentage Difference All points All lung points Points inside lung Points inside lung near the solid water interface All bone points Points inside bone and near the solid water interface Points inside solid water near the bone interface
Standard Deviation
3.13 3.96 2.85 5.19
5.20 5.27 1.69 7.37
1.28 5.02
4.67 2.84
⫺2.71
2.64
A positive value indicates measurement is greater than calculation.
244
Medical Dosimetry
Volume 36, Number 3, 2011
Fig. 6. The ratios of measured to calculated doses at each TLD location for the 6 MV plan. The point locations refer to the labels indicated in Fig. 4.
(Fig. 4). The agreement is good in bone near the bone– solid water interface (0.05 ⫾ 3.03%), but is worse in points in solid water near the bone interface (–7.81 ⫾ 5.23%). Table 2 presents the same results for the 10 MV beam. The agreement between measured and calculated values within the lung (2.85 ⫾ 1.69%) and in points near the lung–solid water interface (5.19 ⫾ 7.37%) is worse than that for the 6 MV beam and shows an underestimation of the dose by the planning system. The agreement in bone near the bone–solid water interface (5.02 ⫾ 2.84%) follows the same trend but is opposite in points in solid water near the bone interface (⫺2.71 ⫾ 2.64%). Figure 5 shows the dose distribution in and around these points. Overall, most measurements are within 5% of the calculated ones with the exception of those near interface and/or in high-dose gradient areas.
The ratios of measured to calculated doses for each TLD location are presented in Figs. 6 and 7 for the 2 energies. The point locations refer to those indicated in Figs. 4 and 5. The error bars represent the propagation of error, summing up the measurement and calculation errors. It is evident from the figures that larger standard deviations are observed in points near the interface and/or high-dose gradient regions. These dose gradients are reflected in large calculation uncertainties as well as increased setup and measurements error caused by both finite size of the dosimeter and increased contribution of setup uncertainties to variations in measurements in these areas. Nevertheless, except for 4 10 MV points at the lung–solid water interface, measurements fall within a single standard deviation. Measured doses in the solid water section of the phantom were also compared with the plan-predicted
Fig. 7. The ratios of measured to calculated doses at each TLD location for the 10 MV plan. The point locations refer to the labels indicated in Fig. 4.
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
ones, and the average percentage difference between measured and computed values is approximately 1% for both energies, which is well within the uncertainty of TLD measurements and indicates there is no systematic difference between measurements for both energies.
8. 9. 10.
DISCUSSION AND CONCLUSIONS 11.
The convolution/superposition algorithm used in the Pinnacle system is known to produce acceptable results in and near inhomogeneities. This study shows the degree of accuracy of this algorithm in realistic cases including regions with low- and high-density inhomogenities, steep dose gradients, and, potentially, electronic disequilibrium. Overall, the accuracy of calculations is good considering the regions of high-dose gradient in which measurements were performed, the finite size of the TLDs, the uncertainty in TLD dosimetry, and the interface effect caused by the introduction of TLD material in lung- or bone-equivalent sections of the phantom. We have found acceptable results (within ⫾5% of measured doses) for the beam energies and inhomogeneities studied here, with comparable results for both 6 and 10 MV, except in the high-gradient, lung–solid water interface region, where one might expect electronic equilibrium to be less well met for the higher energy. The agreement was good in areas within the bone- or lung-equivalent inserts. Overall, we find Pinnacle to adequately model doses in IMRT plans to warrant clinical application, understanding that there will always be greater uncertainty in high-gradient and interface regions.
12. 13.
14. 15.
16. 17. 18.
19.
20.
REFERENCES 21. 1. Mackie, T.R.; El-Khatib, E.; Battista, J.; et al. Lung dose corrections for 6- and 15-MV ⫻ rays. Med. Phys. 12:327–32; 1985. 2. Tang, W.L.; Khan, F.M.; Gerbi, B.J. Validity of lung correction algorithms. Med. Phys. 13:683– 6; 1986. 3. Wong, J.W.; Purdy, J.A. On methods of inhomogeneity corrections for photon transport. Med. Phys. 17:807–14; 1990. 4. Kappas, C.; Rosenwald, J.-C. Quality control of inhomogeneity correction algorithms used in treatment planning systems. Int. J. Radiat. Oncol. Biol. Phys. 32:847–55; 1995. 5. Arnfield, M.R.; Siantar, C.H.; Siebers, J.; et al. The impact of electron transport on the accuracy of computed dose. Med. Phys. 27:1266 –74; 2000. 6. du Plessis, F.C.P.; Willemse, C.A.; Lotter, M.G.; et al. Comparison of the Batho, ETAR and Monte Carlo dose calculation methods in CT based patient models. Med. Phys. 28:582–9; 2001. 7. Martens, C.; Reynaert, N.; De Wagter, C.; et al. Underdosage of the upper-airway mucosa for small fields as used in intensitymodulated radiation therapy: A comparison between radiochromic
22.
23. 24. 25. 26.
AND
P.D. HIGGINS
245
film measurements, Monte Carlo simulations, and collapsed cone convolution calculations. Med. Phys. 29:1528 –35; 2002. Wang, L. Monte Carlo evaluation of 6 MV intensity modulated radiotherapy plans for head and neck and lung treatments. Med. Phys. 29:2705–17; 2002. Jones, A.O.; Das, I.J. A Monte Carlo study of IMRT beamlets in inhomogeneous media. Med. Phys. 30:296 –300; 2003. Jones, A.O.; Das, I.J. Comparison of inhomogeneity correction algorithms in small photon fields. Med. Phys. 32:766 –76; 2005. Boudreau, C.; Heath, E.; Seuntjens, J.; et al. IMRT head and neck treatment planning with a commercially available Monte Carlo based planning system. Phys. Med. Biol. 50:879 –90; 2005. Al-Hallaq, H.A.; Reft, C.S.; Roeske, J.C. The dosimetric effects of tissue heterogeneities in intensity-modulated radiation therapy (IMRT) of the head and neck. Phys. Med. Biol. 51:1145–56; 2006. Vanderstraeten, B.; Rayneart, N.; Paelinck, L.; et al. Accuracy of patient dose calculation for lung IMRT: A comparison of Monte Carlo, convolution/superposition, and pencil beam computations. Med. Phys. 33:3149 –58; 2006. Dvorak, P.; Stock, M.; Kroupa, B.; et al. Analysis of the dose calculation accuracy for IMRT in lung: A 2D approach. Acta. Oncol. 46:928 –36; 2007. Davidson, S.; Ibbott, G.S.; Prado, K.L.; et al. Accuracy of two heterogeneity dose calculation algorithms for IMRT in treatment plans designed using an anthropomorphic phantom. Med. Phys. 34:1850 –7; 2007. Yoon, M.; Lee, D.-H.; Shin, D.; et al. Accuracy of inhomogeneity correction algorithm in intensity-modulated radiotherapy of headand-neck tumors. Med. Dosim. 32:44 –51; 2007. Carrasco, P.; Jornet, N.; Duch, M.A.; et al. Comparison of dose calculation algorithms in slab phantoms with cortical bone equivalent heterogeneities. Med. Phys. 34:3323–33; 2007. Wilcox, E.E.; Daskalov, G.M. Accuracy of dose measurements and calculations within and beyond heterogeneous tissues for 6 MV photon fields smaller than 4 cm produced by Cyberknife. Med. Phys. 35:2259 – 66; 2008. Davidson, S.E.; Popple, R.A.; Ibbott, G.S.; et al. Technical note: Heterogeneity dose calculation accuracy in IMRT: Study of five commercial treatment planning systems using an anthropomorphic thorax phantom. Med. Phys. 35:5434 –9; 2008. Woo, M.K.; Cunningham, J.R. The validity of the density scaling method in primary electron transport for photon and electron beams. Med. Phys. 17:187–94; 1990. Yu, C.X.; Mackie, T.R.; Wong, J.W. Photon dose calculation incorporating explicit electron transport. Med. Phys. 22:1157– 65; 1995. Papanikolaou, N.; Battista, J.J.; Boyer, A.L.; et al. Tissue inhomogeneity corrections for megavoltage photon beams, report of task group 65 of the radiation therapy committee of the American Association of Physicists in Medicine. Madison, WI: Medical Physics Publishing; 2004. Das, I.J.; Ding, G.X.; Ahnesjo, A. Small fields: Nonequilibrium radiation dosimetry. Med. Phys. 35:206 –15; 2007. Saxena, R.; Higgins, P.D. Measurement and evaluation of inhomogeneity corrections and monitor unit verifications for treatment planning. Med. Dosim. 35:19 –27; 2010. Mackie, T.R.; Scrimger, J.W.; Battista, J.J. A convolution method of calculating dose for 15-MV ⫻ rays. Med. Phys. 12:188 –96; 1985. Ahnesjo, A. Collapsed cone convolution of radiant energy for photon dose calculation in heterogeneous media. Med. Phys. 16: 577–92; 1989.
doi:10.1016/j.meddos.2010.03.010
THE ACCURACY OF INHOMOGENEITY CORRECTIONS IN INTENSITY MODULATED RADIATION THERAPY PLANNING IN PHILIPS PINNACLE SYSTEM PARHAM ALAEI, PH.D., and PATRICK D. HIGGINS, PH.D. Department of Therapeutic Radiology-Radiation Oncology, University of Minnesota, Minneapolis, MN (Received 3 February 2010; accepted 29 March 2010)
Abstract—The degree of accuracy of inhomogeneity corrections in a treatment planning system is dependent on the algorithm used by the system. The choice of field size, however, could have an effect on the calculation accuracy as well. There have been several evaluation studies on the accuracy of inhomogeneity corrections used by different algorithms. Most of these studies, however, focus on evaluating the dose in phantom using simplified geometry and open/static fields. This work focuses on evaluating the degree of dose accuracy in calculations involving intensity-modulated radiation therapy (IMRT) fields incident on a phantom containing both lung- and bone-equivalent heterogeneities using 6 and 10 MV beams. IMRT treatment plans were generated using the Philips Pinnacle treatment planning system and delivered to a phantom containing 55 thermoluminescent dosimeter (TLD) locations within the lung and bone and near the lung and bone interfaces with solid water. The TLD readings were compared with the dose predicted by the planning system. We find satisfactory agreement between planned and delivered doses, with an overall absolute average difference between measurement and calculation of 1.2% for the 6 MV and 3.1% for the 10 MV beam with larger variations observed near the interfaces and in areas of high-dose gradient. The results presented here demonstrate that the convolution algorithm used in the Pinnacle treatment planning system produces accurate results in IMRT plans calculated and delivered to inhomogeneous media, even in regions that potentially lack electronic equilibrium. © 2011 American Association of Medical Dosimetrists. Key Words: Heterogeneity corrections, IMRT, Thermoluminescent dosimetry.
INTRODUCTION
Intensity-modulated radiation therapy (IMRT) treatment planning, because of the presence of small fields, introduces new challenges in dose calculation accuracy, both in homogeneous and inhomogeneous media. Conventional treatment planning rarely involves fields smaller than 4 ⫻ 4 cm2, with the exception of radiosurgery, whereas IMRT planning often involves fields smaller than 4 ⫻ 4 cm2, possibly as small as 1 ⫻ 1 cm2, depending on the system settings and/or preferences set by the user. This can lead to conditions of electronic disequilibrium, which may not be accurately accounted for by the treatment planning system.23 The degree to which a treatment planning system can predict the dose accurately for small fields has been evaluated by a few investigators.5,7,10,17,18,24 We have evaluated the accuracy of the Philips Pinnacle3 treatment planning system (Philips Medical Systems, Milpitas, CA) in modeling the dose within and near the interface of inhomogeneities for IMRT plans in phantom. The collapsed cone convolution algorithm employed in Pinnacle25 is known to have a higher degree of accuracy in inhomogeneous media than other algorithms.5,7,23 Ahnesjo26 has shown that collapsed cone convolution can calculate the dose in inhomogeneous media with very high accuracy, provided charged particle equilibrium is present. So, although collapsed cone convolution is one of the most accurate methods for dose
Inhomogeneity corrections are commonly used in computerized treatment planning. Depending on the algorithm, there are varying degrees to which a treatment planning system can accurately predict the dose in and around an inhomogeneity. There are many publications on the evaluation of inhomogeneity corrections in treatment planning systems.1–18 Most of these publications evaluate this in simple geometry, e.g., in slab phantoms with the addition of lung- or bone-equivalent slabs in between polystyrene or solid water slabs.2–5,9,10,14,16 –18 There have also been a few investigations evaluating the dose in humanoid phantoms12,15,16 and a few patient studies.6,8,11,13 The 2 common methods of evaluation compare the planpredicted dose with measurements5,12,15–19 and with Monte Carlo calculations.6 –11,13 Among the different treatment planning algorithms, the ones using pencil-beam algorithms with correctionbased inhomogeneity calculations are known to suffer from varying degrees of inaccuracies in predicting the dose in and around inhomogeneities.7,20 –22 Systems using convolution-based algorithms or Monte Carlo are known to predict the dose more accurately.7,21,23 Reprint requests to: Parham Alaei, Ph.D., Department of Therapeutic Radiology-Radiation Oncology, University of Minnesota, 420 Delaware Street SE, MMC 494, Minneapolis, MN 55455. E-mail: [email protected] 240
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
AND
P.D. HIGGINS
241
finding better accuracy at the lung–soft tissue interfaces than at the bone–soft tissue ones for field sizes as small as 2 ⫻ 2 cm2. This study compares Pinnacle-predicted doses to measurements in a phantom containing both lung- and boneequivalent heterogeneities for intensity-modulated beams. MATERIALS AND METHODS
Fig. 1. The phantom used for the study with the lung and bone inserts shown in the insets. Locations of initial TLD slots and additional milled ones are indicated in the figure. The overall phantom size is 30 cm high by 35 cm wide by 13 cm thick. The lung-equivalent insert has a square cross-section of 6.5 ⫻ 6.5 cm and the bone-equivalent one is a cylinder of diameter 3 cm.
calculation in inhomogeneous materials, this limitation warrants an investigation into the degree of accuracy of inhomogeneity corrections in the Pinnacle system. This can only be evaluated with detailed measurements. Previously, Martens et al.,7 Jones and Das,10 AlHallaq et al.,12 and Davidson et al.15,19 have done comparison studies that included the Pinnacle system. Martens et al.7 showed that although collapsed cone convolution does not predict the dose accurately at the air-phantom interface, it is an improvement over pencil beam algorithms. Jones and Das10 compared the dose in a phantom with low-density heterogeneities predicted by various algorithms, including collapsed cone convolution, with that calculated using Monte Carlo and concluded that collapsed cone convolution models the dose more accurately in lung than other methods evaluated. Al-Hallaq et al.12 compared the dose predicted by the CORVUS planning system (Best NOMOS, Pittsburgh, PA) with thermoluminescent dosimeters (TLDs) placed in the Rando head phantom. They then imported the multileaf collimator (MLC) leaf sequences into Pinnacle and calculated the dose. However, the earlier version of Pinnacle used in that study does not account for MLC leakage and rounded leaf edge. They reported good agreement between measured and Pinnacle-predicted doses. Davidson et al.15 evaluated the accuracy of dose calculations by CORVUS and Pinnacle systems in a thorax phantom and concluded that Pinnacle predicts the dose to within 5% of measured dose. In another paper, Davidson et al.19 evaluated the accuracy of 5 treatment planning systems, including Pinnacle, in an anthropomorphic thorax phantom. Saxena and Higgins24 compared measured and Pinnacle-computed doses in a phantom including lung and bone inhomogeneities for a range of square field sizes,
Phantom The phantom used for evaluation is a modified Benchmark IMRT QA phantom made by Civco (Civco Medical Solutions, Kalona, IA). The phantom includes lung- and bone-equivalent inhomogeneities, with a number of slots for TLD placement in the lung-equivalent insert. To measure the dose within and near these inhomogeneities, additional slots of 4.5-mm diameter and 1-mm depth were milled in the lung-equivalent insert as well as in and near the bone-equivalent insert (Fig. 1). The slots’ dimensions allow for placement of lithium fluoride (LiF) TLD chips, with a minimal amount of air between the chips and the phantom. Treatment planning Pinnacle treatment planning system v8.0m was used to generate IMRT plans on the phantom. The planning system was commissioned for the 6 and 10 MV beams, and modeling was performed, resulting in excellent agreement between modeled and measured profiles for both energies. The agreement between modeled and measured profiles is better than 1% in points beyond buildup region for depth dose curves and within the field for cross profiles. The accuracy of the dose calculation algorithm was tested using simple geometry according to the test described by Tang et al.2 The measurements
Fig. 2. The transverse CT slice of the phantom indicating locations of the target and avoidance structures and the inhomogeneities.
242
Medical Dosimetry
Volume 36, Number 3, 2011
The optimization was done using the Direct Machine Parameter Optimization (DMPO) module of the Pinnacle system and the parameters used were: maximum number of segments—70; minimum segment area—2 cm2; and minimum segment monitor units—5. All optimization and calculations were done heterogeneously with a dose grid of 3 mm3. The calculations were also repeated using a dose grid of 2 mm3 for comparison. A routine IMRT quality assurance (QA), consisting of point dose measurements and film dosimetry, was then performed on the plans. To obtain the expected dose at measurement locations, each TLD was contoured on the corresponding CT slice and the mean and standard deviation of the dose at each TLD position was obtained from the planning system. Fig. 3. The 6 MV IMRT dose distribution within phantom at the CT slice corresponding to the TLD locations within lungequivalent inhomogeneity. The isodose lines are at 10-cGy increments.
were within 1.5% of those reported in this paper for 6 MV and within 2.4% for 10 MV, the larger discrepancies being in the build-up region and at the interfaces. The planning system was also evaluated as part of the work by Saxena and Higgins.24 The phantom was computed tomography (CT)– scanned and 2 7-field IMRT plans were generated by optimizing to a spherical target within the phantom, between the lung and bone inserts. Three additional avoidance structures were added to achieve a realistic IMRT plan with significant modulation (Fig. 2). The inhomogeneities were not used as constraints in the optimization. One plan used 6- and the other used 10 MV beams from a Varian 21ex linear accelerator (Varian Medical Systems, Palo Alto, CA) equipped with 60-pair millennium MLCs.
Dosimeters LiF TLD chips of dimensions 3.2 ⫻ 3.2 ⫻ 0.9 mm were used for dose measurements. A batch of TLDs was calibrated by exposing them to a known dose 3 times, averaging the readings, and calculating the percent difference between each reading and the average of 3 readings. Nitrogen gas flow was used to stabilize the readings. After each irradiation, the chips were left for 24 hours before reading using a Harshaw Model 3500 TLD reader (Thermo Scientific, Waltham, MA) and were post-annealed at 400 °C for 1 hour. The chips with the least amount of variation of response between the 3 measurements (standard deviation of ⬍3%) were chosen and an average calibration factor (cGy/nC) was calculated for each chip. The chips were calibrated for both beam energies and the calibrations were verified throughout the experiment. Measurements The planned IMRT was delivered to phantom, with the calibrated TLD chips placed in a total of 55 locations,
Fig. 4. The dose distribution of the 6 MV plan in phantom within and near the inhomogeneities. Locations of TLD measurement points are indicated, along with the point labels. For the points inside lung-equivalent inhomogeneity (a), there are 6 rows of 5 TLD locations each (points A1–E6) and an additional 2 columns of 4 TLDs each on each side (points a1–a4 and b1– b4). For the points inside and near the bone-equivalent inhomogeneity (b), there is 1 point at the center (0); another 8 points inside the inhomogeneity labeled 1– 8, with number 1 on the top and sequencing clockwise; and another 8 points outside the inhomogeneity labeled 9 –16, with number 9 on top and sequencing clockwise as well.
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
AND
P.D. HIGGINS
243
Fig. 5. The dose distribution for the 10 MV plan in phantom within and near inhomogeneities. Locations of TLD measurements are described in the Fig. 4 legend.
38 inside the lung insert, 18 near the lung–solid water interface, and the remaining 20 inside the lung. Seventeen TLDs were placed in the bone insert and in the surrounding solid water near the bone–solid water interface. All TLDs were placed inside the milled slots in direct contact with phantom material. The phantom was positioned for treatment with the aid of external radioopaque markers placed on it during CT scan. The measurements were performed at least 7 times for each energy/plan. Linear accelerator output during the repeat irradiations was within 0.5% and not accounted for in these measurements. Separate TLD measurements were also performed inside the solid water section of the phantom (upper right of Fig. 1) to evaluate measurement uncertainty far from the interfaces. RESULTS Figure 3 shows the dose distributions within phantom for the 6 MV plan. The contours showing the locations of the TLDs in lung can be seen in the figures. The isodose lines are at 10-cGy increments indicating areas of small- and large-dose gradient. The dose distribution for the 10 MV plan is similar. Figures 4 and 5 are enlarged versions of the 6 and 10 MV plans, respecTable 1. The results of comparison of measured to calculated doses for the IMRT plan using 6 MV beam
All points All lung points Points inside lung Points inside lung near the solid water interface All bone points Points inside bone and near the solid water interface Points inside solid water near the bone interface
Average Percentage Difference
Standard Deviation
⫺1.16 ⫺0.16 ⫺0.14 ⫺0.17
4.98 4.18 2.00 5.91
⫺3.39 0.05
5.96 3.03
⫺7.81
5.23
A positive value indicates the measurement is greater than calculation.
tively, showing the isodose distribution within and near the inhomogeneities and locations of the TLD measurement points. A summary of the TLD measurements for the 6 MV plan can be found in Table 1. The table shows the percentage difference between measured and planned doses at each location. Listed measured doses represent the average of 7 measurements, with corresponding standard deviations. Calculations are the mean dose within each contoured TLD position along with standard deviations as reported by the planning system. The lung measurement points have been divided into points inside the lung (⬎1.5 cm from lung–solid water interface) and points near the solid water interface (⬍1.0 cm from the lung–solid water interface). The bone measurements have been divided into points inside bone and near (⬃0.5 cm) the solid water interface and those in solid water near (⬃0.5 cm) the bone interface. Only one point in the center of the bone insert can be considered far (⬃1.5 cm) from the interface. We find there to be generally good agreement between measured and calculated values within the lung (– 0.14 ⫾ 2.00%), with an increased uncertainty in points near the lung–solid water interface (– 0.17 ⫾ 5.91%), some of which also fall within high-dose gradient areas Table 2. The results of comparison of measured to calculated doses for the IMRT plan using 10 MV beam Average Percentage Difference All points All lung points Points inside lung Points inside lung near the solid water interface All bone points Points inside bone and near the solid water interface Points inside solid water near the bone interface
Standard Deviation
3.13 3.96 2.85 5.19
5.20 5.27 1.69 7.37
1.28 5.02
4.67 2.84
⫺2.71
2.64
A positive value indicates measurement is greater than calculation.
244
Medical Dosimetry
Volume 36, Number 3, 2011
Fig. 6. The ratios of measured to calculated doses at each TLD location for the 6 MV plan. The point locations refer to the labels indicated in Fig. 4.
(Fig. 4). The agreement is good in bone near the bone– solid water interface (0.05 ⫾ 3.03%), but is worse in points in solid water near the bone interface (–7.81 ⫾ 5.23%). Table 2 presents the same results for the 10 MV beam. The agreement between measured and calculated values within the lung (2.85 ⫾ 1.69%) and in points near the lung–solid water interface (5.19 ⫾ 7.37%) is worse than that for the 6 MV beam and shows an underestimation of the dose by the planning system. The agreement in bone near the bone–solid water interface (5.02 ⫾ 2.84%) follows the same trend but is opposite in points in solid water near the bone interface (⫺2.71 ⫾ 2.64%). Figure 5 shows the dose distribution in and around these points. Overall, most measurements are within 5% of the calculated ones with the exception of those near interface and/or in high-dose gradient areas.
The ratios of measured to calculated doses for each TLD location are presented in Figs. 6 and 7 for the 2 energies. The point locations refer to those indicated in Figs. 4 and 5. The error bars represent the propagation of error, summing up the measurement and calculation errors. It is evident from the figures that larger standard deviations are observed in points near the interface and/or high-dose gradient regions. These dose gradients are reflected in large calculation uncertainties as well as increased setup and measurements error caused by both finite size of the dosimeter and increased contribution of setup uncertainties to variations in measurements in these areas. Nevertheless, except for 4 10 MV points at the lung–solid water interface, measurements fall within a single standard deviation. Measured doses in the solid water section of the phantom were also compared with the plan-predicted
Fig. 7. The ratios of measured to calculated doses at each TLD location for the 10 MV plan. The point locations refer to the labels indicated in Fig. 4.
Accuracy of inhomogeneity corrections in IMRT ● P. ALAEI
ones, and the average percentage difference between measured and computed values is approximately 1% for both energies, which is well within the uncertainty of TLD measurements and indicates there is no systematic difference between measurements for both energies.
8. 9. 10.
DISCUSSION AND CONCLUSIONS 11.
The convolution/superposition algorithm used in the Pinnacle system is known to produce acceptable results in and near inhomogeneities. This study shows the degree of accuracy of this algorithm in realistic cases including regions with low- and high-density inhomogenities, steep dose gradients, and, potentially, electronic disequilibrium. Overall, the accuracy of calculations is good considering the regions of high-dose gradient in which measurements were performed, the finite size of the TLDs, the uncertainty in TLD dosimetry, and the interface effect caused by the introduction of TLD material in lung- or bone-equivalent sections of the phantom. We have found acceptable results (within ⫾5% of measured doses) for the beam energies and inhomogeneities studied here, with comparable results for both 6 and 10 MV, except in the high-gradient, lung–solid water interface region, where one might expect electronic equilibrium to be less well met for the higher energy. The agreement was good in areas within the bone- or lung-equivalent inserts. Overall, we find Pinnacle to adequately model doses in IMRT plans to warrant clinical application, understanding that there will always be greater uncertainty in high-gradient and interface regions.
12. 13.
14. 15.
16. 17. 18.
19.
20.
REFERENCES 21. 1. Mackie, T.R.; El-Khatib, E.; Battista, J.; et al. Lung dose corrections for 6- and 15-MV ⫻ rays. Med. Phys. 12:327–32; 1985. 2. Tang, W.L.; Khan, F.M.; Gerbi, B.J. Validity of lung correction algorithms. Med. Phys. 13:683– 6; 1986. 3. Wong, J.W.; Purdy, J.A. On methods of inhomogeneity corrections for photon transport. Med. Phys. 17:807–14; 1990. 4. Kappas, C.; Rosenwald, J.-C. Quality control of inhomogeneity correction algorithms used in treatment planning systems. Int. J. Radiat. Oncol. Biol. Phys. 32:847–55; 1995. 5. Arnfield, M.R.; Siantar, C.H.; Siebers, J.; et al. The impact of electron transport on the accuracy of computed dose. Med. Phys. 27:1266 –74; 2000. 6. du Plessis, F.C.P.; Willemse, C.A.; Lotter, M.G.; et al. Comparison of the Batho, ETAR and Monte Carlo dose calculation methods in CT based patient models. Med. Phys. 28:582–9; 2001. 7. Martens, C.; Reynaert, N.; De Wagter, C.; et al. Underdosage of the upper-airway mucosa for small fields as used in intensitymodulated radiation therapy: A comparison between radiochromic
22.
23. 24. 25. 26.
AND
P.D. HIGGINS
245
film measurements, Monte Carlo simulations, and collapsed cone convolution calculations. Med. Phys. 29:1528 –35; 2002. Wang, L. Monte Carlo evaluation of 6 MV intensity modulated radiotherapy plans for head and neck and lung treatments. Med. Phys. 29:2705–17; 2002. Jones, A.O.; Das, I.J. A Monte Carlo study of IMRT beamlets in inhomogeneous media. Med. Phys. 30:296 –300; 2003. Jones, A.O.; Das, I.J. Comparison of inhomogeneity correction algorithms in small photon fields. Med. Phys. 32:766 –76; 2005. Boudreau, C.; Heath, E.; Seuntjens, J.; et al. IMRT head and neck treatment planning with a commercially available Monte Carlo based planning system. Phys. Med. Biol. 50:879 –90; 2005. Al-Hallaq, H.A.; Reft, C.S.; Roeske, J.C. The dosimetric effects of tissue heterogeneities in intensity-modulated radiation therapy (IMRT) of the head and neck. Phys. Med. Biol. 51:1145–56; 2006. Vanderstraeten, B.; Rayneart, N.; Paelinck, L.; et al. Accuracy of patient dose calculation for lung IMRT: A comparison of Monte Carlo, convolution/superposition, and pencil beam computations. Med. Phys. 33:3149 –58; 2006. Dvorak, P.; Stock, M.; Kroupa, B.; et al. Analysis of the dose calculation accuracy for IMRT in lung: A 2D approach. Acta. Oncol. 46:928 –36; 2007. Davidson, S.; Ibbott, G.S.; Prado, K.L.; et al. Accuracy of two heterogeneity dose calculation algorithms for IMRT in treatment plans designed using an anthropomorphic phantom. Med. Phys. 34:1850 –7; 2007. Yoon, M.; Lee, D.-H.; Shin, D.; et al. Accuracy of inhomogeneity correction algorithm in intensity-modulated radiotherapy of headand-neck tumors. Med. Dosim. 32:44 –51; 2007. Carrasco, P.; Jornet, N.; Duch, M.A.; et al. Comparison of dose calculation algorithms in slab phantoms with cortical bone equivalent heterogeneities. Med. Phys. 34:3323–33; 2007. Wilcox, E.E.; Daskalov, G.M. Accuracy of dose measurements and calculations within and beyond heterogeneous tissues for 6 MV photon fields smaller than 4 cm produced by Cyberknife. Med. Phys. 35:2259 – 66; 2008. Davidson, S.E.; Popple, R.A.; Ibbott, G.S.; et al. Technical note: Heterogeneity dose calculation accuracy in IMRT: Study of five commercial treatment planning systems using an anthropomorphic thorax phantom. Med. Phys. 35:5434 –9; 2008. Woo, M.K.; Cunningham, J.R. The validity of the density scaling method in primary electron transport for photon and electron beams. Med. Phys. 17:187–94; 1990. Yu, C.X.; Mackie, T.R.; Wong, J.W. Photon dose calculation incorporating explicit electron transport. Med. Phys. 22:1157– 65; 1995. Papanikolaou, N.; Battista, J.J.; Boyer, A.L.; et al. Tissue inhomogeneity corrections for megavoltage photon beams, report of task group 65 of the radiation therapy committee of the American Association of Physicists in Medicine. Madison, WI: Medical Physics Publishing; 2004. Das, I.J.; Ding, G.X.; Ahnesjo, A. Small fields: Nonequilibrium radiation dosimetry. Med. Phys. 35:206 –15; 2007. Saxena, R.; Higgins, P.D. Measurement and evaluation of inhomogeneity corrections and monitor unit verifications for treatment planning. Med. Dosim. 35:19 –27; 2010. Mackie, T.R.; Scrimger, J.W.; Battista, J.J. A convolution method of calculating dose for 15-MV ⫻ rays. Med. Phys. 12:188 –96; 1985. Ahnesjo, A. Collapsed cone convolution of radiant energy for photon dose calculation in heterogeneous media. Med. Phys. 16: 577–92; 1989.