- Email: [email protected]
Improvement in clinical step and shoot intensity modulated radiation therapy delivery accuracy on an integrated linear accelerator control system
Practical Radiation Oncology (2014) 4, 43–49
www.practicalradonc.org
Original Report
Improvement in clinical step and shoot intensity modulated radiation therapy delivery accuracy on an integrated linear accelerator control system C.E. Agnew PhD a,⁎, D.M. Irvine PhD a , A.R. Hounsell PhD a, b , C.K. McGarry PhD a, b a
Radiotherapy Physics, Northern Ireland Cancer Centre, Belfast Health and Social Care Trust, Belfast, Northern Ireland, United Kingdom b Centre for Cancer Research and Cell Biology, Queen’s University Belfast, Northern Ireland, United Kingdom Received 1 May 2013; revised 5 June 2013; accepted 1 July 2013
Abstract Purpose: The dose delivery accuracy of 30 clinical step and shoot intensity modulated radiation therapy plans was investigated using the single integrated multileaf collimator controller of the Varian Truebeam linear accelerator (linac) (Varian Medical Systems, Palo Alto, CA) and compared with the dose delivery accuracy on a previous generation Varian 2100CD C-Series linac. Methods and Materials: Ten prostate, 10 prostate and pelvic node, and 10 head-and-neck cases were investigated in this study. Dose delivery accuracy on each linac was assessed using Farmer ionization chamber point dose measurements, 2-dimensional planar ionization chamber array measurements, and the corresponding Varian dynamic log files. Absolute point dose measurements, fluence delivery accuracy, leaf position accuracy, and the overshoot effect were assessed for each plan. Results: Absolute point dose delivery accuracy increased by 1.5% on the Truebeam compared with the 2100CD linac. No improvement in fluence delivery accuracy between the linacs, at a gamma criterion of 3%/3 mm was measured using the 2-dimensional ionization chamber array, with median (interquartile range) gamma passing rates of 98.99% (97.70%-99.72%) and 99.28% (98.26%-99.75%) for the Truebeam and 2100CD linacs, respectively. Varian log files also showed no improvement in fluence delivery between the linacs at 3%/3 mm, with median gamma passing rates of 99.97% (99.93%-99.99%) and 99.98% (99.94%-100%) for the Truebeam and 2100CD linacs, respectively. However, log files revealed improved leaf position accuracy and fluence delivery at 1%/1 mm criterion on the Truebeam (99.87%; 99.78%-99.94%) compared with the 2100CD linac (97.87%; 91.93%-99.49%). The overshoot effect, characterized on the 2100CD linac, was not observed on the Truebeam. Conclusions: The integrated multileaf collimator controller on the Varian Truebeam improves clinical treatment delivery accuracy of step and shoot intensity modulated radiation therapy fields compared with delivery on a Varian C-series linac. Crown Copyright © 2014 Published by Elsevier Inc. on behalf of American Society for Radiation Oncology. All rights reserved.
Conflicts of interest: None. ⁎ Corresponding author. Belfast Health and Social Care Trust, Radiotherapy Physics, NICC, Belfast, BT9 7AB, Northern Ireland, United Kingdom. E-mail address: [email protected] (C.E. Agnew). 1879-8500/$ – see front matter. Crown Copyright © 2014 Published by Elsevier Inc. on behalf of American Society for Radiation Oncology. All rights reserved. http://dx.doi.org/10.1016/j.prro.2013.07.003
44
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
Introduction High delivery accuracy is required for intensity modulated radiation therapy (IMRT) due to the highly conformal and complex nature of IMRT plans. Due to potential errors in IMRT plan dose calculations, plan transfer, and beam delivery, it is routine for all patient plans to undergo patient-specific quality control (QC). 1 IMRT delivery verification has been conventionally carried out using phantom-based measurements, 2 but more recently the use of phantom-less techniques have been reported. 3-6 Phantom-based methods include Farmer ionization chamber, 2-dimensional (2D) and 3-dimensional ionization chamber or diode arrays and film measurements, 2 while phantom-less measurements include trajectory log file analysis 3-5 and electronic portal imaging device dosimetry. 6 The accuracy and precision of the treatment machine delivery relies on the synchronization of the multileaf collimator (MLC) controller and the linear accelerator (linac) beam delivery control system. 7 In the Varian Cseries linacs (Varian Medical Systems, Palo Alto, CA), communication between these 2 control systems takes approximately 50 to 80 ms. 8 For step and shoot IMRT (SIMRT) deliveries this time delay between the MLC controller and the linac controller monitoring and halting the beam delivery results in a phenomenon known as the overshoot effect. 8 The overshoot effect results in the first segment being consistently overdosed, while the dose to intermediate segments varies. Dose to the final segment is consistently undershot, as the beam is terminated by the linac control system, independent from the MLC controller, when the total monitor unit (MU) is reached. Alternatively, the new generation Varian Truebeam linacs (Varian Medical Systems) have an integrated MLC and linac control system. This integrated system communicates with each component retrospectively every 10 ms and uses this information to prospectively instruct each component for the subsequent 10 ms and 20 ms in order to synchronize the planned and actual treatment delivery. The integrated
Table 1
control system of the Truebeam has been found to improve the dose delivery accuracy of SIMRT fields, particularly for low dose segments (1 or 2 MU), at high dose rates of up to 600 MU/minute, with no obvious overshoot or undershoot trend. 7,9 Work to date has assessed the delivery accuracy of the Truebeam integrated control system on geometric fields using phantom-based measurements. In this work, we assess the delivery accuracy of 30 clinical plans on Varian Truebeam and Varian 2100CD linacs using both phantom-based measurements (Farmer ionization chamber and 2D ionization chamber array) and Varian log files. Phantom-based measurements enable quantification of absolute dose delivery and an independent assessment of fluence delivery accuracy. Varian log file analysis permits assessment of MLC position accuracy, fractional dose delivery accuracy, fluence delivery accuracy, and direct quantification of the overshoot effect.
Methods and materials Ten prostate only, 10 prostate and pelvic node (PPN), and 10 head-and-neck (H&N) SIMRT plans, previously accepted for clinical treatment, were retrospectively assessed in this study. All plans were generated on Oncentra V3.3 and V4.1 (Nucletron BV, Veenendal, The Netherlands), using 6 MV photon beams and 400 MU/ minute dose rate, in accordance with the current clinical implementation of SIMRT in the Northern Ireland Cancer Centre (NICC). During commissioning of the treatment planning system, the focal spot size and MLC leakage were optimized based on data measured in the NICC. 10 A summary of patient plan statistics is provided in Table 1. The modulation complexity score (MCS), which assess the complexity of a plan based on the variability of leaf positions and aperture areas between segments, 11 was determined for each plan and the average score for each treatment site is presented in Table 1. Patient-specific QC measurements were used to assess the delivery accuracy of each patient plan on 2 Varian
Patient plan demographics
Variable
Prostate 12
PPN 13
H&N 14
No. of patients No. of fields Field gantry angles (degrees)
10 5 180, 100, 35, 325, 260 74 Gy in 37 77.1 50 6 (1-43)
10 5 180, 95, 35, 325, 265 74 Gy in 37 256.8 50 9 (2-44)
10 7 206, 257.5, 309, 51.5, 103, 154.5, 265 70 Gy in 35 257.7 70 9 (1-40)
0.63 ± 0.09
0.43 ± 0.06
0.35 ± 0.10
Fractionation Average field size (cm 2) Segments MU per segment median, (min-max range) MCS (mean ± 1 SD) 11
H&N, head and neck; PPN, prostate and pelvic node.
Practical Radiation Oncology: January-February 2014
Improvement SIMRT delivery on Truebeam
Truebeam linacs and 4 Varian 2100CD linacs. The 6 MV photon beams on all 6 linacs were found to be dosimetrically equivalent to within ± 1% along percentage depth dose curves and profiles. All linacs were equipped with a 120 Millennium MLC (Varian Medical Systems).
Phantom-based measurements Point dose measurements Per field point dose delivery accuracy was determined using a Farmer ionization chamber measurement positioned isocentrically in a 20 × 30 × 30 cm 3 water equivalent phantom (Barts and The London NHS Trust, London, UK). The Farmer chamber was traceably calibrated to the National Physics Laboratory primary standard using the 1990 code of practice for high energy photon therapy dosimetry. 15 All Farmer ionization chamber measurements were delivered at the planned gantry angle. All treatment plans were recalculated in Oncentra V3.3 and V4.1 (Elekta, Stockholm, Sweden) on a computed tomography scan of the Farmer ionization chamber positioned within 20 cm of 30 × 30 cm 2 solid water. The Farmer chamber (FC) measurements were compared with the average planned dose to the Farmer chamber volume as detailed in Eq (1). Point Dose Accuracy ¼
Measured FC Dose Planned FC Dose
ð1Þ
45
log files were converted into ASCII format and also assessed using the in-house MATLAB software. In addition to the information contained within the C-series DynaLog files, the Truebeam trajectory log files also provide details of the couch position, dose rate, beam energy, and absolute MUs delivered. 16 Additionally, the Truebeam log files record the beam delivery parameters every 20 ms (50 Hz) rather than the 50 ms (20 Hz) used in the DynaLog files. MLC position errors Leaf position errors for each MLC leaf bank (A/B) ΔXA/B were determined at each log file sample, as described in Eq (2), by comparing the actual delivered log file leaf positions (ΔXA/Bdelivered) to the planned leaf positions (ΔXA/Bplanned) as detailed in the RT DICOM (Digital Imaging and Communications in Medicine) plan. DX A=B ¼ X A=B planned − X A=B delivered
ð2Þ
The average (Av), the root mean square (RMS), and the standard deviation (SD) of leaf position errors were used to summarize the leaf position errors from each log file sample. Errors were averaged over all segments for each field as previously described. 5 Histograms of absolute leaf position errors at each sampled interval were also assessed for each field. All results included errors during the overshoot effect. The software was designed to only consider open, in-field leaf positions within each segment.
Fluence delivery accuracy Per field fluence delivery accuracy was determined using a 2D ionization chamber array (MatriXX; IBA Dosimetry, Schwarzenbruck, Germany) positioned with 30 × 30 × 10 cm 3 water equivalent phantom backscatter and 30 × 30 × 4.5 cm 3 build-up to provide an equivalent ionization chamber depth of 5 cm. Each field was delivered to the MatriXX at gantry angle 0 degrees. The MatriXX was calibrated in absolute dose on each linac. Treatment plans were recalculated on a 30 × 30 × 20 cm 3 solid water phantom and the coronal slice at depth 5 cm used for comparison with the MatriXX measurements. The 2D fluence delivery was assessed with gamma analysis using OmniPro 1.6 analysis software (IBA Dosimetry) with a 3%/3 mm gamma criterion.
Dose delivery errors Delivered fluence maps were reconstructed using the log file actual MLC leaf positions and the log file actual fractional MUs delivered per segment. Fluence maps were constructed as the relative sum of all segments. The actual delivered fluence was compared with the 2D planned fluence map reconstructed equivalently, using the delivery parameters contained in the RT DICOM plan. Gamma criterion of 3%/3 mm and 1%/1 mm were used to assess fluence delivery accuracy. To assess the overshoot effect, the error in dose delivery to the first and last segment of each field was determined as a percentage of the planned segment dose as described in Eq (3).
Phantom-less Based Measurements
%Segment Dose Errori Delivered Dose Segi − Planned Dose Segi ð3Þ ¼100 ; Planned Dose Segi
Varian dynamic log files (DynaLogs and trajectory logs) recorded for each field during the Farmer ionization chamber measurements at planned gantry angles and 2D ionization chamber array measurements at gantry angle 0 degrees were assessed, totaling 3196 log files. Analysis was carried out as previously described 5 using in-house software written in MATLAB V7.7.0 (MathWorks, Natick, MA). The binary Varian Truebeam trajectory
where i is either the first or last segment.
Statistical analysis Results are presented as mean ± 1 SD or median (interquartile range). Statistical analysis was carried out using SPSS V20.0.0 (SPSS Inc, Chicago, IL). Independent
46
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
sample t tests and, where appropriate, Mann–Whitney U tests, were used to assess differences in the results between Truebeam and 2100CD linacs. Results were deemed significant at P b .05.
Results Point dose measurements Farmer ionization chamber point dose measurements acquired on Truebeam (0.999 ± 0.024) were not significantly different from planned doses (P b .454); ie, close to unity for all treatment sites, as illustrated in Fig 1. However, point dose measurements acquired on 2100CD linacs were significantly different from planned doses (1.015 ± 0.024, P b .001), for all treatment sites as presented in Fig 1. The larger variability in H&N measurements (shown in Fig 1) was noted on both linac types and may be a consequence of the higher complexity of H&N treatment plans (MCS, 0.35 ± 0.10).
MatriXX 2D ionization chamber array MatriXX 2D ionization chamber array measurements on both Truebeam and 2100CD linacs were similar over all treatment sites as illustrated in Fig 2 (P = .92), with an overall passing rate of 98.99% (97.70%-99.72%) and 99.28% (98.26%-99.75%) for the Truebeam and 2100CD linacs, respectively. A relationship between fluence
Figure 2 MatriXX fluence delivery accuracy at 3%/3 mm gamma criterion on Truebeam compared with 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck. Box plots represent the median and interquartile range, with whiskers representing the spread of data.
delivery accuracy and plan complexity was revealed with less complex prostate only treatments (MCS, 0.62 ± 0.09), having a 1% higher gamma passing rate compared with more complex PPN (MCS, 0.43 ± 0.06) and H&N deliveries (MCS, 0.35 ± 0.10).
Log file results
Figure 1 Ratio of Farmer ionization chamber measurement compared with planned point dose measurements acquired on Truebeam and 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck.
MLC position errors The MLC position errors were assessed from 2036 log file pairs. The RMS position errors were significantly smaller (P b .001) for all treatment sites on the Truebeam (0.023 ± 0.015 mm) compared with the 2100CD linacs (0.239 ± 0.066 mm) as presented in Fig 3A. Average leaf positions errors were also significantly smaller (P b .001) for all treatment sites on the Truebeam (0.002 ± 0.004 mm) compared with the 2100CD linacs (0.005 ± 0.008 mm), as presented in Fig 3B. The variability and stability of leaf positioning was assessed by determining the SD of the error in each leaf position at each sampled interval throughout the treatment delivery. Leaf position variability was significantly reduced (P b .001) over all treatment sites on the Truebeam (0.023 ± 0.017 mm) compared with the 2100CD linacs (0.246 ± 0.067 mm). The histograms of absolute leaf position errors revealed Truebeam had no leaf position errors N 0.1 cm for any treatment delivery, while 2100CD linacs had 1.65% of leaf position errors N 0.1 cm and no leaf position errors N 0.4 cm.
Practical Radiation Oncology: January-February 2014
Improvement SIMRT delivery on Truebeam
47
The influence of the overshoot effect on the delivery of the first and final segments was assessed as described by Eq (3). The results are presented in Table 2. The characteristic overshoot in the first segment and undershoot in the last segment are noted for the 2100CD deliveries. 8 However, no observable overshoot or undershoot was measured during Truebeam deliveries.
Figure 3 (A) Root mean square and (B) average leaf position errors on Truebeam and 2100CD linacs over all treatment sites.
Dose delivery accuracy Gamma analysis at 3%/3 mm criterion revealed both linac types were comparable (P = .277) with a gamma passing rate of 99.97% (99.93%-99.99%) and 99.98% (99.94%-100%) for the Truebeam and 2100CD linacs, respectively (Fig 4A). A gamma criterion of 1%/1 mm was required to reveal differences in the delivery accuracy between the 2 linac types (P b .001), as illustrated in Fig 4B. The Truebeam maintained a high 1%/1 mm gamma passing rate of 99.87% (99.78%-99.94%), while the 2100CD linacs 1%/1 mm gamma passing rate reduced to 97.87% (91.93%-99.49%).
Figure 4 Log file fluence delivery accuracy at (A) 3%/3 mm and (B) 1%/1 mm gamma criterion on Truebeam compared with 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck. Box plots represent the median and interquartile range, with whiskers representing the spread of data.
48
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
Table 2 Quantification of the % overshoot and undershoot in the first and last segment on the Truebeam and 2100CD linac Segment
Truebeam
2100 CD linac
First segment (%) Last segment (%)
0.08 ± 0.13 0.01 ± 0.49
4.69 ± 3.37 − 6.60 ± 7.12
Discussion The Truebeam, with its integrated MLC and linac control system, improved leaf position and dose delivery accuracy for all treatment plans investigated when compared with the 2100CD linacs and its independent MLC and linac controllers. The overshoot effect, which has been a particular issue for SIMRT deliveries on Varian C-series linacs, resulting in the redistribution and omission of MU segments, was found in this work to be negligible during clinical Truebeam SIMRT deliveries, despite the increased sampling rate of 50 Hz. This is in agreement with previous studies of plans with geometric fields using low MU segments. 7,9 A number of studies have suggested a reduction in the overshoot effect, demonstrated through optimal segment sequencing, could improve delivery accuracy by 1%-2%. 5,17 This 1%-2% improvement in delivery accuracy due to a reduction in the overshoot effect is corroborated in this work, where Truebeam Farmer ionization chamber point dose measurements were not significantly different from planned doses, but 2100CD linac point dose measurements were significantly different from planned doses by on average 1.5%. The effect of the increased rate of communication and the predictive rather than reactive communication between the Truebeam MLC and linac components can also be seen in the reduction of Truebeam leaf position errors in comparison with 2100CD leaf position errors. The RMS leaf position errors were reduced on the Truebeam. This is likely to be due to the reduction in the overshoot effect, as leaf movement to the subsequent segment during the overshoot period as seen in the 2100CD deliveries does not occur in Truebeam deliveries. Average leaf position errors are not affected by the overshoot effect 5 and were comparable for both linacs. Of note, average leaf position errors were smaller than previously described due to the removal of in-field closed leaf position errors. 5,18,19 The repeatability and stability of Truebeam deliveries was also significantly improved in comparison with 2100CD linacs as indicated by a reduction in SD of leaf positioning errors, gamma passing rates, and with the elimination of the overshoot effect. Combining improvements in dose delivery and leaf position accuracy in the Truebeam was not found to significantly improve fluence delivery accuracy at 3%/ 3 mm gamma criterion between the 2 linacs as assessed using both the 2D ionization chamber array (P = .092) and Varian log files (P = .189). However, a tighter
gamma criterion of 1%/1 mm of log file fluence revealed improvements in the gamma passing rates on Truebeam linacs compared with 2100CD linacs (P b 0.01). DynaLog 1%/1 mm gamma passing rates on 2100CD linacs (97.87%; 91.93%-99.49%) were within the confidence intervals for passing rates previously reported. 5 On Truebeam linacs, the fluence delivery accuracy at 1%/1 mm increased for all treatment sites to 99.87% (99.78%-99.94%). Improvement at 1%/1 mm level in fluence delivery is in agreement with improvements noted in absolute dose delivery measured in this work and in previous studies that reported elimination of the overshoot effect could improve delivery accuracy by 1% to 2%. 5,17 Although not clinically significant, understanding the impact and minimizing each individual uncertainty in the delivery process can enable optimization of treatment delivery. This may be particularly relevant with the move toward delivering more conformal dose distributions, with dose escalation and dose painting. It was not possible to replicate the tighter 1%/ 1 mm criterion on the ionization chamber array measurements due to inherent uncertainties in setting up the device and limitations of detector spacing. 20 Thus, this study also reveals the utility of using Varian log files for QC measurements as the disparity between the Farmer ionization chamber results and the 2D planar array results could not have been explained without the additional level of detail provided by the Varian log files. However, log files do not provide a complete QC solution as they are recorded by the linac and as such are not a true independent measure of plan delivery quality.
Conclusions The integrated MLC and linac control system in the Truebeam significantly reduced leaf position errors and eliminated the overshoot effect, resulting in improved fluence delivery accuracy as compared with the 2100CD linac. Phantom-based Farmer ionization chamber measurements revealed a 1.5% improvement in absolute point dose delivery accuracy on the Truebeam compared with 2100CD linacs. Further investigation of Varian Log files at a gamma criterion of 1%/1 mm replicated the improvements seen in absolute dose measurements, with fluence delivery accuracy at this 1%/1 mm level increasing median gamma passing rates to 99.87% (99.78%-99.94%) on Truebeam compared with 97.87% (91.93%-99.49%) on 2100CD linacs.
References 1. Miles EA, Clark CH, Urbano MT, et al. The impact of introducing intensity modulated radiotherapy. Radiother Oncol. 2005;77:241-246. 2. Ezzell GA, Burmeister JW, Dogan N, et al. IMRT commissioning: Multiple institution planning and dosimetry comparisons, a report from AAPM Task Group 119. Med Phys. 2009;36:5359-5373.
Practical Radiation Oncology: January-February 2014 3. Litzenberg DW, Moran JM, Fraass BA. Verification of dynamic and segmental IMRT delivery by dynamic log file analysis. J Appl Clin Med Phys. 2002;3:63-72. 4. Ranagaraj D, Zhu M, Yang D, et al. Catching errors with patientspecific pretreatment machine log file analysis. Pract Radiat Oncol. 2013;3:80-90. 5. Agnew CE, King RB, Hounsell AR, McGarry CK. Implementation of phantom-less IMRT delivery verification using Varian DynaLog files and R/V output. Phys Med Biol. 2012;57:6761-6777. 6. Wendling M, Louwe RJ, McDermott LN, Sonke JJ, van Herk M, Mijnheer BJ. Accurate two-dimensional IMRT verification. Med Phys. 2006;33:259-273. 7. Li J, Wiersma RD, Stepaniak CJ, Farrey KJ, Al-Hallaq HA. Improvements in dose accuracy delivered with static-MLC IMRT on an integrated linear accelerator control system. Med Phys. 2012;39: 2456-2462. 8. Ezzell GA, Chungbin S. The overshoot phenomenon in step and shoot IMRT delivery. J Appl Clin Med Phys. 2001;2:138-148. 9. Dolan J, Ambrose R, Furhang E, et al. A comparison of small MU sub-field dosimetry for step and shoot IMRT fields on Varian IX and Truebeam machines. [Abstract]. AAPM 54th Annual Meeting 2012; SU-E-T-368. 10. Mayles WPM, Lake R, McKenzie A, et al. IPEM Report 81: Physic aspects of quality control in radiotherapy. The Institute of Physics and Engineering in Medicine, York, England, 1999. 11. McNiven AL, Sharpe MB, Purdie TG. A new metric for assessing IMRT modulation complexity and plan deliverability. Med Phys. 2010;37:505-515. 12. Dearnaley D, Syndikus I, Sumo G, et al. Conventional versus hypofractionated high-dose intensity-modulated radiotherapy for
Improvement SIMRT delivery on Truebeam
13.
14.
15.
16. 17.
18.
19.
20.
49
prostate cancer: preliminary safety results from the CHHiP randomised controlled trial. Lancet Oncol. 2012;13:43-54. Harris V, South C, Cruickshank C, et al. A national phase III trial of pelvic lymph node (LN) IMRT in prostate cancer (PIVOTAL): a comparison of LN outlining methods [Abstract]. Radiat Oncol. 2011;99:S486-S487. Miah AB, Bhide SA, Guerrero-Urbano MT, et al. Dose-escalated intensity-modulated radiotherapy is feasible and may improve locoregional control and laryngeal preservation in laryngohypopharyngeal cancers. Int J Radiat Oncol Biol Phys. 2012;18: 539-547. Lillicrap SC, Owen B, Williams JR, et al. Code of Practice for highenergy photon therapy dosimetry based on the National Physics Laboratory absorbed dose calibration service. Phys Med Biol. 1990;35:1355-1360. Varian Medical Systems, Truebeam Trajectory Log File Specification 100049068–02, USA, 2011. Kuperman VY, Lam WC. Improving delivery of segments with small MU in step and shoot IMRT. Med Phys. 2006;33:10671073. Luo W, Li J, Price RA Jr, et al. Monte Carlo based IMRT dose verification using MLC log files and R/V outputs. Med Phys. 2006;33:2557-2564. Sun B, Rangaraj D, Palaniswaamy G, et al. Initial experience with Truebeam trajectory log files for radiation therapy delivery verification. Practical Radiat Oncol. 2013; [In Press] Available, online 15 January 2013. Poppe B, Blechschmidt A, Djouguela A, et al. Two-dimensional ionization chamber arrays for IMRT plan verification. Med Phys. 2006;33:1005-1015.
www.practicalradonc.org
Original Report
Improvement in clinical step and shoot intensity modulated radiation therapy delivery accuracy on an integrated linear accelerator control system C.E. Agnew PhD a,⁎, D.M. Irvine PhD a , A.R. Hounsell PhD a, b , C.K. McGarry PhD a, b a
Radiotherapy Physics, Northern Ireland Cancer Centre, Belfast Health and Social Care Trust, Belfast, Northern Ireland, United Kingdom b Centre for Cancer Research and Cell Biology, Queen’s University Belfast, Northern Ireland, United Kingdom Received 1 May 2013; revised 5 June 2013; accepted 1 July 2013
Abstract Purpose: The dose delivery accuracy of 30 clinical step and shoot intensity modulated radiation therapy plans was investigated using the single integrated multileaf collimator controller of the Varian Truebeam linear accelerator (linac) (Varian Medical Systems, Palo Alto, CA) and compared with the dose delivery accuracy on a previous generation Varian 2100CD C-Series linac. Methods and Materials: Ten prostate, 10 prostate and pelvic node, and 10 head-and-neck cases were investigated in this study. Dose delivery accuracy on each linac was assessed using Farmer ionization chamber point dose measurements, 2-dimensional planar ionization chamber array measurements, and the corresponding Varian dynamic log files. Absolute point dose measurements, fluence delivery accuracy, leaf position accuracy, and the overshoot effect were assessed for each plan. Results: Absolute point dose delivery accuracy increased by 1.5% on the Truebeam compared with the 2100CD linac. No improvement in fluence delivery accuracy between the linacs, at a gamma criterion of 3%/3 mm was measured using the 2-dimensional ionization chamber array, with median (interquartile range) gamma passing rates of 98.99% (97.70%-99.72%) and 99.28% (98.26%-99.75%) for the Truebeam and 2100CD linacs, respectively. Varian log files also showed no improvement in fluence delivery between the linacs at 3%/3 mm, with median gamma passing rates of 99.97% (99.93%-99.99%) and 99.98% (99.94%-100%) for the Truebeam and 2100CD linacs, respectively. However, log files revealed improved leaf position accuracy and fluence delivery at 1%/1 mm criterion on the Truebeam (99.87%; 99.78%-99.94%) compared with the 2100CD linac (97.87%; 91.93%-99.49%). The overshoot effect, characterized on the 2100CD linac, was not observed on the Truebeam. Conclusions: The integrated multileaf collimator controller on the Varian Truebeam improves clinical treatment delivery accuracy of step and shoot intensity modulated radiation therapy fields compared with delivery on a Varian C-series linac. Crown Copyright © 2014 Published by Elsevier Inc. on behalf of American Society for Radiation Oncology. All rights reserved.
Conflicts of interest: None. ⁎ Corresponding author. Belfast Health and Social Care Trust, Radiotherapy Physics, NICC, Belfast, BT9 7AB, Northern Ireland, United Kingdom. E-mail address: [email protected] (C.E. Agnew). 1879-8500/$ – see front matter. Crown Copyright © 2014 Published by Elsevier Inc. on behalf of American Society for Radiation Oncology. All rights reserved. http://dx.doi.org/10.1016/j.prro.2013.07.003
44
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
Introduction High delivery accuracy is required for intensity modulated radiation therapy (IMRT) due to the highly conformal and complex nature of IMRT plans. Due to potential errors in IMRT plan dose calculations, plan transfer, and beam delivery, it is routine for all patient plans to undergo patient-specific quality control (QC). 1 IMRT delivery verification has been conventionally carried out using phantom-based measurements, 2 but more recently the use of phantom-less techniques have been reported. 3-6 Phantom-based methods include Farmer ionization chamber, 2-dimensional (2D) and 3-dimensional ionization chamber or diode arrays and film measurements, 2 while phantom-less measurements include trajectory log file analysis 3-5 and electronic portal imaging device dosimetry. 6 The accuracy and precision of the treatment machine delivery relies on the synchronization of the multileaf collimator (MLC) controller and the linear accelerator (linac) beam delivery control system. 7 In the Varian Cseries linacs (Varian Medical Systems, Palo Alto, CA), communication between these 2 control systems takes approximately 50 to 80 ms. 8 For step and shoot IMRT (SIMRT) deliveries this time delay between the MLC controller and the linac controller monitoring and halting the beam delivery results in a phenomenon known as the overshoot effect. 8 The overshoot effect results in the first segment being consistently overdosed, while the dose to intermediate segments varies. Dose to the final segment is consistently undershot, as the beam is terminated by the linac control system, independent from the MLC controller, when the total monitor unit (MU) is reached. Alternatively, the new generation Varian Truebeam linacs (Varian Medical Systems) have an integrated MLC and linac control system. This integrated system communicates with each component retrospectively every 10 ms and uses this information to prospectively instruct each component for the subsequent 10 ms and 20 ms in order to synchronize the planned and actual treatment delivery. The integrated
Table 1
control system of the Truebeam has been found to improve the dose delivery accuracy of SIMRT fields, particularly for low dose segments (1 or 2 MU), at high dose rates of up to 600 MU/minute, with no obvious overshoot or undershoot trend. 7,9 Work to date has assessed the delivery accuracy of the Truebeam integrated control system on geometric fields using phantom-based measurements. In this work, we assess the delivery accuracy of 30 clinical plans on Varian Truebeam and Varian 2100CD linacs using both phantom-based measurements (Farmer ionization chamber and 2D ionization chamber array) and Varian log files. Phantom-based measurements enable quantification of absolute dose delivery and an independent assessment of fluence delivery accuracy. Varian log file analysis permits assessment of MLC position accuracy, fractional dose delivery accuracy, fluence delivery accuracy, and direct quantification of the overshoot effect.
Methods and materials Ten prostate only, 10 prostate and pelvic node (PPN), and 10 head-and-neck (H&N) SIMRT plans, previously accepted for clinical treatment, were retrospectively assessed in this study. All plans were generated on Oncentra V3.3 and V4.1 (Nucletron BV, Veenendal, The Netherlands), using 6 MV photon beams and 400 MU/ minute dose rate, in accordance with the current clinical implementation of SIMRT in the Northern Ireland Cancer Centre (NICC). During commissioning of the treatment planning system, the focal spot size and MLC leakage were optimized based on data measured in the NICC. 10 A summary of patient plan statistics is provided in Table 1. The modulation complexity score (MCS), which assess the complexity of a plan based on the variability of leaf positions and aperture areas between segments, 11 was determined for each plan and the average score for each treatment site is presented in Table 1. Patient-specific QC measurements were used to assess the delivery accuracy of each patient plan on 2 Varian
Patient plan demographics
Variable
Prostate 12
PPN 13
H&N 14
No. of patients No. of fields Field gantry angles (degrees)
10 5 180, 100, 35, 325, 260 74 Gy in 37 77.1 50 6 (1-43)
10 5 180, 95, 35, 325, 265 74 Gy in 37 256.8 50 9 (2-44)
10 7 206, 257.5, 309, 51.5, 103, 154.5, 265 70 Gy in 35 257.7 70 9 (1-40)
0.63 ± 0.09
0.43 ± 0.06
0.35 ± 0.10
Fractionation Average field size (cm 2) Segments MU per segment median, (min-max range) MCS (mean ± 1 SD) 11
H&N, head and neck; PPN, prostate and pelvic node.
Practical Radiation Oncology: January-February 2014
Improvement SIMRT delivery on Truebeam
Truebeam linacs and 4 Varian 2100CD linacs. The 6 MV photon beams on all 6 linacs were found to be dosimetrically equivalent to within ± 1% along percentage depth dose curves and profiles. All linacs were equipped with a 120 Millennium MLC (Varian Medical Systems).
Phantom-based measurements Point dose measurements Per field point dose delivery accuracy was determined using a Farmer ionization chamber measurement positioned isocentrically in a 20 × 30 × 30 cm 3 water equivalent phantom (Barts and The London NHS Trust, London, UK). The Farmer chamber was traceably calibrated to the National Physics Laboratory primary standard using the 1990 code of practice for high energy photon therapy dosimetry. 15 All Farmer ionization chamber measurements were delivered at the planned gantry angle. All treatment plans were recalculated in Oncentra V3.3 and V4.1 (Elekta, Stockholm, Sweden) on a computed tomography scan of the Farmer ionization chamber positioned within 20 cm of 30 × 30 cm 2 solid water. The Farmer chamber (FC) measurements were compared with the average planned dose to the Farmer chamber volume as detailed in Eq (1). Point Dose Accuracy ¼
Measured FC Dose Planned FC Dose
ð1Þ
45
log files were converted into ASCII format and also assessed using the in-house MATLAB software. In addition to the information contained within the C-series DynaLog files, the Truebeam trajectory log files also provide details of the couch position, dose rate, beam energy, and absolute MUs delivered. 16 Additionally, the Truebeam log files record the beam delivery parameters every 20 ms (50 Hz) rather than the 50 ms (20 Hz) used in the DynaLog files. MLC position errors Leaf position errors for each MLC leaf bank (A/B) ΔXA/B were determined at each log file sample, as described in Eq (2), by comparing the actual delivered log file leaf positions (ΔXA/Bdelivered) to the planned leaf positions (ΔXA/Bplanned) as detailed in the RT DICOM (Digital Imaging and Communications in Medicine) plan. DX A=B ¼ X A=B planned − X A=B delivered
ð2Þ
The average (Av), the root mean square (RMS), and the standard deviation (SD) of leaf position errors were used to summarize the leaf position errors from each log file sample. Errors were averaged over all segments for each field as previously described. 5 Histograms of absolute leaf position errors at each sampled interval were also assessed for each field. All results included errors during the overshoot effect. The software was designed to only consider open, in-field leaf positions within each segment.
Fluence delivery accuracy Per field fluence delivery accuracy was determined using a 2D ionization chamber array (MatriXX; IBA Dosimetry, Schwarzenbruck, Germany) positioned with 30 × 30 × 10 cm 3 water equivalent phantom backscatter and 30 × 30 × 4.5 cm 3 build-up to provide an equivalent ionization chamber depth of 5 cm. Each field was delivered to the MatriXX at gantry angle 0 degrees. The MatriXX was calibrated in absolute dose on each linac. Treatment plans were recalculated on a 30 × 30 × 20 cm 3 solid water phantom and the coronal slice at depth 5 cm used for comparison with the MatriXX measurements. The 2D fluence delivery was assessed with gamma analysis using OmniPro 1.6 analysis software (IBA Dosimetry) with a 3%/3 mm gamma criterion.
Dose delivery errors Delivered fluence maps were reconstructed using the log file actual MLC leaf positions and the log file actual fractional MUs delivered per segment. Fluence maps were constructed as the relative sum of all segments. The actual delivered fluence was compared with the 2D planned fluence map reconstructed equivalently, using the delivery parameters contained in the RT DICOM plan. Gamma criterion of 3%/3 mm and 1%/1 mm were used to assess fluence delivery accuracy. To assess the overshoot effect, the error in dose delivery to the first and last segment of each field was determined as a percentage of the planned segment dose as described in Eq (3).
Phantom-less Based Measurements
%Segment Dose Errori Delivered Dose Segi − Planned Dose Segi ð3Þ ¼100 ; Planned Dose Segi
Varian dynamic log files (DynaLogs and trajectory logs) recorded for each field during the Farmer ionization chamber measurements at planned gantry angles and 2D ionization chamber array measurements at gantry angle 0 degrees were assessed, totaling 3196 log files. Analysis was carried out as previously described 5 using in-house software written in MATLAB V7.7.0 (MathWorks, Natick, MA). The binary Varian Truebeam trajectory
where i is either the first or last segment.
Statistical analysis Results are presented as mean ± 1 SD or median (interquartile range). Statistical analysis was carried out using SPSS V20.0.0 (SPSS Inc, Chicago, IL). Independent
46
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
sample t tests and, where appropriate, Mann–Whitney U tests, were used to assess differences in the results between Truebeam and 2100CD linacs. Results were deemed significant at P b .05.
Results Point dose measurements Farmer ionization chamber point dose measurements acquired on Truebeam (0.999 ± 0.024) were not significantly different from planned doses (P b .454); ie, close to unity for all treatment sites, as illustrated in Fig 1. However, point dose measurements acquired on 2100CD linacs were significantly different from planned doses (1.015 ± 0.024, P b .001), for all treatment sites as presented in Fig 1. The larger variability in H&N measurements (shown in Fig 1) was noted on both linac types and may be a consequence of the higher complexity of H&N treatment plans (MCS, 0.35 ± 0.10).
MatriXX 2D ionization chamber array MatriXX 2D ionization chamber array measurements on both Truebeam and 2100CD linacs were similar over all treatment sites as illustrated in Fig 2 (P = .92), with an overall passing rate of 98.99% (97.70%-99.72%) and 99.28% (98.26%-99.75%) for the Truebeam and 2100CD linacs, respectively. A relationship between fluence
Figure 2 MatriXX fluence delivery accuracy at 3%/3 mm gamma criterion on Truebeam compared with 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck. Box plots represent the median and interquartile range, with whiskers representing the spread of data.
delivery accuracy and plan complexity was revealed with less complex prostate only treatments (MCS, 0.62 ± 0.09), having a 1% higher gamma passing rate compared with more complex PPN (MCS, 0.43 ± 0.06) and H&N deliveries (MCS, 0.35 ± 0.10).
Log file results
Figure 1 Ratio of Farmer ionization chamber measurement compared with planned point dose measurements acquired on Truebeam and 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck.
MLC position errors The MLC position errors were assessed from 2036 log file pairs. The RMS position errors were significantly smaller (P b .001) for all treatment sites on the Truebeam (0.023 ± 0.015 mm) compared with the 2100CD linacs (0.239 ± 0.066 mm) as presented in Fig 3A. Average leaf positions errors were also significantly smaller (P b .001) for all treatment sites on the Truebeam (0.002 ± 0.004 mm) compared with the 2100CD linacs (0.005 ± 0.008 mm), as presented in Fig 3B. The variability and stability of leaf positioning was assessed by determining the SD of the error in each leaf position at each sampled interval throughout the treatment delivery. Leaf position variability was significantly reduced (P b .001) over all treatment sites on the Truebeam (0.023 ± 0.017 mm) compared with the 2100CD linacs (0.246 ± 0.067 mm). The histograms of absolute leaf position errors revealed Truebeam had no leaf position errors N 0.1 cm for any treatment delivery, while 2100CD linacs had 1.65% of leaf position errors N 0.1 cm and no leaf position errors N 0.4 cm.
Practical Radiation Oncology: January-February 2014
Improvement SIMRT delivery on Truebeam
47
The influence of the overshoot effect on the delivery of the first and final segments was assessed as described by Eq (3). The results are presented in Table 2. The characteristic overshoot in the first segment and undershoot in the last segment are noted for the 2100CD deliveries. 8 However, no observable overshoot or undershoot was measured during Truebeam deliveries.
Figure 3 (A) Root mean square and (B) average leaf position errors on Truebeam and 2100CD linacs over all treatment sites.
Dose delivery accuracy Gamma analysis at 3%/3 mm criterion revealed both linac types were comparable (P = .277) with a gamma passing rate of 99.97% (99.93%-99.99%) and 99.98% (99.94%-100%) for the Truebeam and 2100CD linacs, respectively (Fig 4A). A gamma criterion of 1%/1 mm was required to reveal differences in the delivery accuracy between the 2 linac types (P b .001), as illustrated in Fig 4B. The Truebeam maintained a high 1%/1 mm gamma passing rate of 99.87% (99.78%-99.94%), while the 2100CD linacs 1%/1 mm gamma passing rate reduced to 97.87% (91.93%-99.49%).
Figure 4 Log file fluence delivery accuracy at (A) 3%/3 mm and (B) 1%/1 mm gamma criterion on Truebeam compared with 2100CD linacs grouped into treatment site; prostate, prostate and pelvic node, and head and neck. Box plots represent the median and interquartile range, with whiskers representing the spread of data.
48
C.E. Agnew et al
Practical Radiation Oncology: January-February 2014
Table 2 Quantification of the % overshoot and undershoot in the first and last segment on the Truebeam and 2100CD linac Segment
Truebeam
2100 CD linac
First segment (%) Last segment (%)
0.08 ± 0.13 0.01 ± 0.49
4.69 ± 3.37 − 6.60 ± 7.12
Discussion The Truebeam, with its integrated MLC and linac control system, improved leaf position and dose delivery accuracy for all treatment plans investigated when compared with the 2100CD linacs and its independent MLC and linac controllers. The overshoot effect, which has been a particular issue for SIMRT deliveries on Varian C-series linacs, resulting in the redistribution and omission of MU segments, was found in this work to be negligible during clinical Truebeam SIMRT deliveries, despite the increased sampling rate of 50 Hz. This is in agreement with previous studies of plans with geometric fields using low MU segments. 7,9 A number of studies have suggested a reduction in the overshoot effect, demonstrated through optimal segment sequencing, could improve delivery accuracy by 1%-2%. 5,17 This 1%-2% improvement in delivery accuracy due to a reduction in the overshoot effect is corroborated in this work, where Truebeam Farmer ionization chamber point dose measurements were not significantly different from planned doses, but 2100CD linac point dose measurements were significantly different from planned doses by on average 1.5%. The effect of the increased rate of communication and the predictive rather than reactive communication between the Truebeam MLC and linac components can also be seen in the reduction of Truebeam leaf position errors in comparison with 2100CD leaf position errors. The RMS leaf position errors were reduced on the Truebeam. This is likely to be due to the reduction in the overshoot effect, as leaf movement to the subsequent segment during the overshoot period as seen in the 2100CD deliveries does not occur in Truebeam deliveries. Average leaf position errors are not affected by the overshoot effect 5 and were comparable for both linacs. Of note, average leaf position errors were smaller than previously described due to the removal of in-field closed leaf position errors. 5,18,19 The repeatability and stability of Truebeam deliveries was also significantly improved in comparison with 2100CD linacs as indicated by a reduction in SD of leaf positioning errors, gamma passing rates, and with the elimination of the overshoot effect. Combining improvements in dose delivery and leaf position accuracy in the Truebeam was not found to significantly improve fluence delivery accuracy at 3%/ 3 mm gamma criterion between the 2 linacs as assessed using both the 2D ionization chamber array (P = .092) and Varian log files (P = .189). However, a tighter
gamma criterion of 1%/1 mm of log file fluence revealed improvements in the gamma passing rates on Truebeam linacs compared with 2100CD linacs (P b 0.01). DynaLog 1%/1 mm gamma passing rates on 2100CD linacs (97.87%; 91.93%-99.49%) were within the confidence intervals for passing rates previously reported. 5 On Truebeam linacs, the fluence delivery accuracy at 1%/1 mm increased for all treatment sites to 99.87% (99.78%-99.94%). Improvement at 1%/1 mm level in fluence delivery is in agreement with improvements noted in absolute dose delivery measured in this work and in previous studies that reported elimination of the overshoot effect could improve delivery accuracy by 1% to 2%. 5,17 Although not clinically significant, understanding the impact and minimizing each individual uncertainty in the delivery process can enable optimization of treatment delivery. This may be particularly relevant with the move toward delivering more conformal dose distributions, with dose escalation and dose painting. It was not possible to replicate the tighter 1%/ 1 mm criterion on the ionization chamber array measurements due to inherent uncertainties in setting up the device and limitations of detector spacing. 20 Thus, this study also reveals the utility of using Varian log files for QC measurements as the disparity between the Farmer ionization chamber results and the 2D planar array results could not have been explained without the additional level of detail provided by the Varian log files. However, log files do not provide a complete QC solution as they are recorded by the linac and as such are not a true independent measure of plan delivery quality.
Conclusions The integrated MLC and linac control system in the Truebeam significantly reduced leaf position errors and eliminated the overshoot effect, resulting in improved fluence delivery accuracy as compared with the 2100CD linac. Phantom-based Farmer ionization chamber measurements revealed a 1.5% improvement in absolute point dose delivery accuracy on the Truebeam compared with 2100CD linacs. Further investigation of Varian Log files at a gamma criterion of 1%/1 mm replicated the improvements seen in absolute dose measurements, with fluence delivery accuracy at this 1%/1 mm level increasing median gamma passing rates to 99.87% (99.78%-99.94%) on Truebeam compared with 97.87% (91.93%-99.49%) on 2100CD linacs.
References 1. Miles EA, Clark CH, Urbano MT, et al. The impact of introducing intensity modulated radiotherapy. Radiother Oncol. 2005;77:241-246. 2. Ezzell GA, Burmeister JW, Dogan N, et al. IMRT commissioning: Multiple institution planning and dosimetry comparisons, a report from AAPM Task Group 119. Med Phys. 2009;36:5359-5373.
Practical Radiation Oncology: January-February 2014 3. Litzenberg DW, Moran JM, Fraass BA. Verification of dynamic and segmental IMRT delivery by dynamic log file analysis. J Appl Clin Med Phys. 2002;3:63-72. 4. Ranagaraj D, Zhu M, Yang D, et al. Catching errors with patientspecific pretreatment machine log file analysis. Pract Radiat Oncol. 2013;3:80-90. 5. Agnew CE, King RB, Hounsell AR, McGarry CK. Implementation of phantom-less IMRT delivery verification using Varian DynaLog files and R/V output. Phys Med Biol. 2012;57:6761-6777. 6. Wendling M, Louwe RJ, McDermott LN, Sonke JJ, van Herk M, Mijnheer BJ. Accurate two-dimensional IMRT verification. Med Phys. 2006;33:259-273. 7. Li J, Wiersma RD, Stepaniak CJ, Farrey KJ, Al-Hallaq HA. Improvements in dose accuracy delivered with static-MLC IMRT on an integrated linear accelerator control system. Med Phys. 2012;39: 2456-2462. 8. Ezzell GA, Chungbin S. The overshoot phenomenon in step and shoot IMRT delivery. J Appl Clin Med Phys. 2001;2:138-148. 9. Dolan J, Ambrose R, Furhang E, et al. A comparison of small MU sub-field dosimetry for step and shoot IMRT fields on Varian IX and Truebeam machines. [Abstract]. AAPM 54th Annual Meeting 2012; SU-E-T-368. 10. Mayles WPM, Lake R, McKenzie A, et al. IPEM Report 81: Physic aspects of quality control in radiotherapy. The Institute of Physics and Engineering in Medicine, York, England, 1999. 11. McNiven AL, Sharpe MB, Purdie TG. A new metric for assessing IMRT modulation complexity and plan deliverability. Med Phys. 2010;37:505-515. 12. Dearnaley D, Syndikus I, Sumo G, et al. Conventional versus hypofractionated high-dose intensity-modulated radiotherapy for
Improvement SIMRT delivery on Truebeam
13.
14.
15.
16. 17.
18.
19.
20.
49
prostate cancer: preliminary safety results from the CHHiP randomised controlled trial. Lancet Oncol. 2012;13:43-54. Harris V, South C, Cruickshank C, et al. A national phase III trial of pelvic lymph node (LN) IMRT in prostate cancer (PIVOTAL): a comparison of LN outlining methods [Abstract]. Radiat Oncol. 2011;99:S486-S487. Miah AB, Bhide SA, Guerrero-Urbano MT, et al. Dose-escalated intensity-modulated radiotherapy is feasible and may improve locoregional control and laryngeal preservation in laryngohypopharyngeal cancers. Int J Radiat Oncol Biol Phys. 2012;18: 539-547. Lillicrap SC, Owen B, Williams JR, et al. Code of Practice for highenergy photon therapy dosimetry based on the National Physics Laboratory absorbed dose calibration service. Phys Med Biol. 1990;35:1355-1360. Varian Medical Systems, Truebeam Trajectory Log File Specification 100049068–02, USA, 2011. Kuperman VY, Lam WC. Improving delivery of segments with small MU in step and shoot IMRT. Med Phys. 2006;33:10671073. Luo W, Li J, Price RA Jr, et al. Monte Carlo based IMRT dose verification using MLC log files and R/V outputs. Med Phys. 2006;33:2557-2564. Sun B, Rangaraj D, Palaniswaamy G, et al. Initial experience with Truebeam trajectory log files for radiation therapy delivery verification. Practical Radiat Oncol. 2013; [In Press] Available, online 15 January 2013. Poppe B, Blechschmidt A, Djouguela A, et al. Two-dimensional ionization chamber arrays for IMRT plan verification. Med Phys. 2006;33:1005-1015.