- Email: [email protected]
Evaluation of functionally weighted dose-volume parameters for thoracic stereotactic ablative radiotherapy (SABR) using CT ventilation
Physica Medica 49 (2018) 47–51
Contents lists available at ScienceDirect
Physica Medica journal homepage: www.elsevier.com/locate/ejmp
Original paper
Evaluation of functionally weighted dose-volume parameters for thoracic stereotactic ablative radiotherapy (SABR) using CT ventilation
T
⁎
Takayuki Kanaia,b, Noriyuki Kadoyaa, , Yujiro Nakajimaa,c, Yuya Miyasakaa,b, Yoshiro Iekoa,b, Tomohiro Kajikawaa, Kengo Itoa, Takaya Yamamotoa, Suguru Dobashid, Ken Takedad, Kenji Nemotob, Keiichi Jingua a
Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan Department of Radiation Oncology, Yamagata University Faculty of Medicine, Yamagata, Japan c Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Tokyo, Japan d Department of Radiological Technology, School of Health Sciences, Faculty of Medicine, Tohoku University, Sendai, Japan b
A R T I C LE I N FO
A B S T R A C T
Keywords: Functional imaging CT ventilation Radiation pneumonitis Functional planning Thoracic cancer
For the purpose of reducing radiation pneumontisis (RP), four-dimensional CT (4DCT)-based ventilation can be used to reduce functionally weighted lung dose. This study aimed to evaluate the functionally weighted dosevolume parameters and to investigate an optimal weighting method to realize effective planning optimization in thoracic stereotactic ablative radiotherapy (SABR). Forty patients treated with SABR were analyzed. Ventilation images were obtained from 4DCT using deformable registration and Hounsfield unit-based calculation. Functionally-weighted mean lung dose (fMLD) and functional lung fraction receiving at least x Gy (fVx) were calculated by two weighting methods: thresholding and linear weighting. Various ventilation thresholds (5th-95th, every 5th percentile) were tested. The predictive accuracy for CTCAE grade ≧ 2 pneumonitis was evaluated by area under the curve (AUC) of receiver operating characteristic analysis. AUC values varied from 0.459 to 0.570 in accordance with threshold and dose-volume metrics. A combination of 25th percentile threshold and fV30 showed the best result (AUC: 0.570). AUC values with fMLD, fV10, fV20, and fV40 were 0.541, 0.487, 0.548 and 0.563 using a 25th percentile threshold. Although conventional MLD, V10, V20, V30 and V40 showed lower AUC values (0.516, 0.477, 0.534, 0.552 and 0.527), the differences were not statistically significant. fV30 with 25th percentile threshold was the best predictor of RP. Our results suggested that the appropriate weighting should be used for better treatment outcomes in thoracic SABR.
1. Introduction CT ventilation is a state-of-the-art imaging modality using four-dimensional computed tomography (4DCT) and deformable image registration [1]. Its clinical validation studies have been done by comparing other ventilation modalities, such as nuclear medicine [2–6], Xenon-CT [7], hyperpolarized 3He-MRI [8], and pulmonary function test [3,9]. These studies showed reasonable correlations between CT ventilation and other modalities. Because 4DCT has already been used in clinical routine and high resolution ventilation maps can easily be calculated only by image processing, this technique is more suitable for high-precision radiotherapy than the other ventilation modalities. These advantages have strongly motivated currently undergoing clinical trials (NCT02843568, NCT02308709, NCT02528942). These trials ⁎
have been aimed for preservation of post-treatment lung function [10,11] and reduction of radiation pneumonitis (RP) [12–15] by avoiding irradiation to highly functioning lung region [16]. However, different optimization methods were used in each research group. There are two major types of reported optimization strategy to consider inhomogeneity of lung function. The first utilizes the threshold to divide functioning and non-functioning area. Yamamoto et al. [12] used 33th and 66th percentile to divide the total lung into three equal volumes: high, moderate, and low-functional lung. They set optimization constraints in each region. Kadoya et al. [15] used 90th percentile for thresholding the functional lung for planning optimization. The second one is linear weighting. Yamamoto et al. [17] and Kida et al. [18] converted ventilation values into percentile and set weight values linearly to the percentile ventilation. Functionally
Corresponding author at: Department of Radiation Oncology, Tohoku University Graduate School of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai 980-8574, Japan. E-mail address: [email protected] (N. Kadoya).
https://doi.org/10.1016/j.ejmp.2018.05.001 Received 23 March 2018; Received in revised form 27 April 2018; Accepted 1 May 2018 Available online 08 May 2018 1120-1797/ © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Physica Medica 49 (2018) 47–51
T. Kanai et al.
Deformable image registration was applied between maximum-inhale CT image HUin(x) and the maximum-exhale image HUex(x). We then calculated the deformation vector u(x) and deformed inhale image HUdeformed(x + u(x)) in each voxel x. The registration process was performed by the elastix [23]: open source registration toolkit. We used a previously reported parameter setting (parameter 2 in [24]) because the mean registration error for lung tissue was reasonably small (< 1.3 mm). All registration results were visually inspected to assure registration accuracy. A Gaussian filter smoothing with a variance σ2 = 1.5 mm2 was subsequently applied to reduce the influence of CT noise [5]. Ventilation value was calculated by Hounsfield unit-based metric (VHU) [1] as:
weighted dose distribution was calculated by the product of weight value and physical dose and used for input to an inverse optimization process. The latter strategy has also been used to correlate functionally weighted dose-volume parameters with incidence of pneumonitis [19]. The optimal weighting strategy which should be used for clinical implementation has been studied by Faught et al. [20] They studied 70 patients treated with conventionally fractionated radiotherapy (CFRT) and concluded that the thresholding-based approach with 84th percentile was the most correlated to grade 3 and higher RP. However, there are no studies which report the correlation between RP and functional dose-volume metrics in thoracic stereotactic ablative radiotherapy (SABR). SABR is characterized as highly conformal dose distribution and hypo-fractionation. These characteristics are known to affect the dose–response of RP[21]. To the best of our knowledge, this is the first study which attempts to quantify clinical significance of CT ventilation in a SABR cohort. We performed receiver operating characteristics (ROC) analysis and evaluated a predictive accuracy of RP by area under the curve (AUC). We also investigated the dependency of AUC values with different weighting approach; thresholding and linear-weighting.
VHU (x) = 1000
W(x,v) =
2.1. Patients and images
The predictive accuracy of grade 2 or greater pneumonitis (CTCAE version 4.0) following radiotherapy was assessed by AUC value of ROC analysis. AUC value is a quantitative value which can assess predictive accuracy of particular indices. AUC value has been used in many studies [19,25,26] because the different prediction from different dose-volume parameters can be easily understood. ROC analysis was performed by in-house MATLAB (The MathWorks Inc., Massachusetts, USA) software. This analysis enabled us to quantify how well different kinds of dosevolume parameters can predict the incidence of radiation pneumonitis. Additionally, the correlation between dose-volume parameters and pneumonitis was also evaluated by univariate analysis. Equality of variance and normality had been confirmed by F-test and Shapiro-Wilk test. Student’s t test was then applied to assess statistical significance with a significance level of 0.05. Statistical analysis was performed by JMP® Pro ver. 12.2.0 (SAS Institute Inc., North Carolina, USA).
For a correction of different fractionation schemes, physical dose distributions were firstly converted into biologically equivalent dose in 2 Gy/fraction (EQD2) using linear-quadratic (LQ) model-based calculation with α / β = 3 [22]. Table 1 Patients’ characteristics and incidence of radiation pneumonitis (RP).
32 8
RP Grade 0 Grade 1 Grade 2
18 13 9
Dose prescription (Gy) Total dose Dose per fraction
40–72 (median 40) 3–10 (median 10)
(2)
2.3. ROC analysis and statistical analysis
2.2. Calculation of functionally weighted dose-volume parameters
58–89 (median 77)
v 50
Above equation means that the highest functioning lung is twice as sensitive to radiation pneumonitis. We obtained weighted dose distribution in each voxel by the product of weight function and dose. Functionally-weighted mean lung dose (fMLD) and functional lung fraction receiving at least 10, 20, 30 and 40 Gy (fV10, fV20, fV30 and fV40) were calculated by the weighted dose distribution. In thresholding method, 19 different percentile thresholds (from 5th to 95th, every 5th percentile) were tested. Functional dose-volume metrics (fMLD, fV10, fV20, fV30 and fV40) were calculated only within voxels which have higher ventilation values than those thresholds.
This study was a retrospective single-institution analysis approved by our institutional review board. Between February 2013 and August 2016, all patients who underwent a pre-treatment 4DCT scan and subsequent thoracic SABR were included. Patients with the following conditions were excluded: (1) History of thoracic radiotherapy, (2) Concurrent chemotherapy, and (3) Follow-up period less than 90 days. In total, 40 patients were analyzed. Patients’ characteristics are summarized in Table 1. Incidence of pneumonitis was evaluated by Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. 4DCT images were acquired as part of clinical therapy to determine internal margin using GE Light Speed RT16 (GE Healthcare, Wisconsin, USA) and Real-time Position Management System (Varian Medical Systems, California, USA). X ray tube current and voltage were 120 mA and 120 kV, respectively. The matrix size in the axial plane was 512 × 512. The axial field of view was determined to include patients’ body. Slice thickness was 2.5 mm. 4DCT images were reconstructed in ten respiratory phases using Advantage 4D workstation (GE Healthcare, Wisconsin, USA)
Age (y)
(1)
To follow other previous studies [12,15,18,19], every ventilation value was converted into percentile. The lowest and the highest ventilation value corresponded to 0th percentile and 100th percentile, respectively. We tested linear weighting method and thresholding method. In linear weighting method, the weight function W(x,v) in each voxel (x) was calculated by percentile ventilation (v) as:
2. Materials and methods
Gender Male Female
HUdeformed ( x+ u(x))−HUex (x) +1 HUex (x){HUdeformed ( x+ u(x)) + 1000}
3. Results Median follow-up period was 14.8 months (range: 3.5–36.9 months). Grade 0, 1, and 2 pneumonitis were observed in 18, 13, and 9 patients, respectively. Fig. 1 shows the summary of AUC values with different methods; thresholding method, linear weighting method and conventional dose-volume parameters. Higher AUC value indicates higher predictive accuracy of the dose-volume parameter. A combination of fV30 and 25th percentile threshold showed the highest predictive accuracy (AUC: 0.570). Twenty-fifth percentile threshold also resulted in good AUC value with fMLD (0.541), fV20 (0.548) and fV40 (0.563), while lower AUC value with fV10 (0.487). These AUC values were higher than those with conventional dose-volume 48
Physica Medica 49 (2018) 47–51
T. Kanai et al.
Fig. 1. Summary of AUC values.
percentile threshold and fV10 with 5th percentile threshold). The difference of AUC values was 0.11. In other words, risk of pneumonitis in 11% of patients can be misestimated by inappropriate choice of dosevolume parameters and weighting method. Therefore, appropriate weighting technique should be used in functional radiotherapy planning. It should be noted that the difference did not reach statistically significant level (p = 0.24).
parameters (0.516, 0.477, 0.534, 0.552 and 0.527 with MLD, V10, V20, V30 and V40, respectively). However, the differences of AUC values were small (ΔAUC < 0.04). Most of the functional dose-volume parameters with linear weighting method showed lower AUC values (0.484, 0.516, 0.520, 0.538 and 0.534 with fMLD, fV10, fV20, fV30 and fV40) than those with conventional dose-volume parameters. The lowest AUC value (0.459) was found in a combination of fV10 and 5th percentile threshold. Any functional or conventional dose-volume parameters did not reach statistically significant level (p > 0.28). Dose distribution and percentile ventilation maps for two representative cases are shown in Fig. 2. In this figure, patient 1 had slightly higher dose-volume parameters (MLD = 5.32 Gy) than patient 2 (MLD = 4.65 Gy). However, high dose areas corresponded to low ventilated region in patient 1 and high ventilated region in patient 2. As a result, functionally-weighted dose-volume parameters with 25th percentile threshold decreased in patient 1 (fMLD = 4.16 Gy) who was not suffered from pneumonitis, and increased in patient 2 (fMLD = 5.29 Gy) who developed grade 2 pneumonitis. Fig. 3 shows the best and the worst ROC curves (fV30 with 25th
4. Discussion We evaluated the predictive accuracy of various functional or conventional dose-volume parameters in SABR. According to Fig. 1, twentieth to thirtieth percentile threshold showed high AUC values, while 70th–80th showed low AUC values. On the contrast, Faught et al. [20] reported 84th percentile resulted in the highest AUC value in CFRT. As compared to CFRT, the lower ventilated lung may contribute the incidence of radiation pneumonitis in hypo-fractionation scheme. In many cases of our results, V30 was more predictive than V20, which commonly used in conventionally fractionated radiotherapy
Fig. 2. Dose distribution and percentile ventilation maps for example cases. 49
Physica Medica 49 (2018) 47–51
T. Kanai et al.
parameters varied with weighting techniques, functional thresholds and dosimetric thresholds. fV30 with 25th percentile would be a good predictor of RP. Our results are useful for functionally optimized planning in SABR. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References [1] Guerrero T, Sanders K, Castillo E, et al. Dynamic ventilation imaging from fourdimensional computed tomography. Phys Med Biol 2006;51:777–91. [2] Castillo R, Castillo E, McCurdy M, et al. Spatial correspondence of 4D CT ventilation and SPECT pulmonary perfusion defects in patients with malignant airway stenosis. Phys Med Biol 2012;57:1855–71. [3] Yamamoto T, Kabus S, Lorenz C, et al. Pulmonary ventilation imaging based on 4dimensional computed tomography: comparison with pulmonary function tests and SPECT ventilation images. Int J Radiat Oncol Biol Phys 2014;90:414–22. [4] Vinogradskiy Y, Koo PJ, Castillo R, et al. Comparison of 4-dimensional computed tomography ventilation with nuclear medicine ventilation-perfusion imaging: a clinical validation study. Int J Radiat Oncol Biol Phys 2014;89:199–205. [5] Kanai T, Kadoya N, Ito K, et al. Evaluation of four-dimensional computed tomography (4D-CT)-based pulmonary ventilation: the high correlation between 4D-CT ventilation and (81m)Kr-planar images was found. Radiother Oncol 2016;119:444–8. [6] Kipritidis J, Siva S, Hofman MS, et al. Validating and improving CT ventilation imaging by correlating with ventilation 4D-PET/CT using 68Ga-labeled nanoparticles. Med Phys 2014;41:011910. [7] Zhang GG, Latifi K, Du K, et al. Evaluation of the DeltaV 4D CT ventilation calculation method using in vivo xenon CT ventilation data and comparison to other methods. J Appl Clin Med Phys 2016;17:550–60. [8] Mathew L, Wheatley A, Castillo R, et al. Hyperpolarized (3)He magnetic resonance imaging: comparison with four-dimensional x-ray computed tomography imaging in lung cancer. Acad Radiol 2012;19:1546–53. [9] Brennan D, Schubert L, Diot Q, et al. Clinical validation of 4-dimensional computed tomography ventilation with pulmonary function test data. Int J Radiat Oncol Biol Phys 2015;92:423–9. [10] Shioyama Y, Jang SY, Liu HH, et al. Preserving functional lung using perfusion imaging and intensity-modulated radiation therapy for advanced-stage non-small cell lung cancer. Int J Radiat Oncol Biol Phys 2007;68:1349–58. [11] Ohashi T, Takeda A, Shigematsu N, et al. Differences in pulmonary function before vs. 1 year after hypofractionated stereotactic radiotherapy for small peripheral lung tumors. Int J Radiat Oncol Biol Phys 2005;62:1003–8. [12] Yamamoto T, Kabus S, von Berg J, et al. Impact of four-dimensional computed tomography pulmonary ventilation imaging-based functional avoidance for lung cancer radiotherapy. Int J Radiat Oncol Biol Phys 2011;79:279–88. [13] Faught AM, Miyasaka Y, Kadoya N, et al. Evaluating the toxicity reduction with computed tomographic ventilation functional avoidance radiation therapy. Int J Radiat Oncol Biol Phys 2017;99:325–33. [14] Waxweiler T, Schubert L, Diot Q, et al. A complete 4DCT-ventilation functional avoidance virtual trial: developing strategies for prospective clinical trials. J Appl Clin Med Phys 2017. [15] Kadoya N, Cho SY, Kanai T, et al. Dosimetric impact of 4-dimensional computed tomography ventilation imaging-based functional treatment planning for stereotactic body radiation therapy with 3-dimensional conformal radiation therapy. Pract Radiat Oncol 2015;5:e505–12. [16] Doi Y, Kimura T, Nakashima T, et al. Functional image-guided radiotherapy planning in volumetric modulated arc therapy for patients with malignant pleural mesothelioma. Adv Radiat Oncol 2017. [17] Yamamoto T, Kabus S, Bal M, et al. The first patient treatment of computed tomography ventilation functional image-guided radiotherapy for lung cancer. Radiother Oncol 2016;118:227–31. [18] Kida S, Bal M, Kabus S, et al. CT ventilation functional image-based IMRT treatment plans are comparable to SPECT ventilation functional image-based plans. Radiother Oncol 2016;118:521–7. [19] Vinogradskiy Y, Castillo R, Castillo E, et al. Use of 4-dimensional computed tomography-based ventilation imaging to correlate lung dose and function with clinical outcomes. Int J Radiat Oncol Biol Phys 2013;86:366–71. [20] Faught AM, Yamamoto T, Castillo R, et al. Evaluating which dose-function metrics are most critical for functional-guided radiation therapy. Int J Radiat Oncol Biol Phys 2017;99:202–9. [21] Borst GR, Ishikawa M, Nijkamp J, et al. Radiation pneumonitis in patients treated for malignant pulmonary lesions with hypofractionated radiation therapy. Radiother Oncol 2009;91:307–13. [22] Borst GR, Ishikawa M, Nijkamp J, et al. Radiation pneumonitis after hypofractionated radiotherapy: evaluation of the LQ(L) model and different dose parameters. Int J Radiat Oncol Biol Phys 2010;77:1596–603. [23] Klein S, Staring M, Murphy K, et al. Elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging 2010;29:196–205. [24] Kanai T, Kadoya N, Ito K, et al. Evaluation of accuracy of B-spline transformation-
Fig. 3. Examples of ROC curves.
[25,27]. Borst et al. [22] studied radiation pneumonitis in hypo-fractionated radiotherapy and reported the best fitting of normal tissue complication probability curve using high dosimetric thresholds, such as 50 Gy. Our results also suggested that higher dosimetric threshold than 20 Gy would be appropriate for prediction of radiation pneumonitis in SABR. Some clinical trials using CT ventilation (NCT02308709, NCT02843568) includes SABR patients as well as CFRT. However, differences between SABR and CFRT in functional planning have yet to be investigated. Our results suggested that the optimal functional dosevolume parameters in SABR may be different from those of CFRT. Although there is a limitation of small sample size, this study is informative for further clinical trials using CT ventilation in a SABR cohort. In this study, physical dose distributions were calculated by treatment planning systems used in our clinical routine. However, the simulated dose distributions in clinically used treatment planning systems are inaccurate in out-of-field area [28] and does not consider the mis-calibration of leaf, gantry, and collimator [29,30]. It should be noted that we ignored these minor errors between simulated and actual dose distributions. CT ventilation images can be affected by the error in deformable image registration [31,32]. Therefore, we used the previously verified registration algorithm and the parameter setting [24]. The registration algorithm and parameter setting are important factors for the accuracy of deformable image registration. Further improvement could be possible by improved registration algorithms [33,34] or adjusting the parameter setting [35]. Biological property of hypo-fractionated radiotherapy is fundamentally different from CFRT. We used LQ model-based calculation to combine different fractionations because in vivo clinical studies showed LQ model is valid to calculate biological equivalent dose even for hypofractionated radiotherapy [22]. However, some cell experiments have suggested inaccuracy of LQ model in high dose per fraction [36]. Some more complex but sophisticated models, such as modified LQ model [37] or lethal–potentially lethal model [38], may be able to improve prediction for radiation pneumonitis in hypo-fractionated irradiation. 5. Conclusion Predictive
accuracy
of
functionally
weighted
dose-volume 50
Physica Medica 49 (2018) 47–51
T. Kanai et al.
[25]
[26]
[27]
[28] [29]
[30]
[31] Castillo E, Castillo R, Vinogradskiy Y, et al. The numerical stability of transformation-based CT ventilation. Int J Comput Assist Radiol Surg 2017. [32] Yamamoto T, Kabus S, Klinder T, et al. Four-dimensional computed tomography pulmonary ventilation images vary with deformable image registration algorithms and metrics. Med Phys 2011;38:1348–58. [33] Samavati N, Velec M, Brock K. A hybrid biomechanical intensity based deformable image registration of lung 4DCT. Phys Med Biol 2015;60:3359–73. [34] Sarrut D, Baudier T, Ayadi M, et al. Deformable image registration applied to lung SBRT: usefulness and limitations. Phys Med 2017;44:108–12. [35] Miura H, Ozawa S, Nakao M, et al. Impact of deformable image registration accuracy on thoracic images with different regularization weight parameter settings. Phys Med 2017;42:108–11. [36] Puck TT, Morkovin D, Marcus PI, et al. Action of x-rays on mammalian cells. II. Survival curves of cells from normal human tissues. J Exp Med 1957;106:485–500. [37] Guerrero M, Li XA. Extending the linear-quadratic model for large fraction doses pertinent to stereotactic radiotherapy. Phys Med Biol 2004;49:4825–35. [38] Curtis SB. Lethal and potentially lethal lesions induced by radiation–a unified repair model. Radiat Res 1986;106:252–70.
based deformable image registration with different parameter settings for thoracic images. J Radiat Res 2014;55:1163–70. Tsujino K, Hirota S, Endo M, et al. Predictive value of dose-volume histogram parameters for predicting radiation pneumonitis after concurrent chemoradiation for lung cancer. Int J Radiat Oncol Biol Phys 2003;55:110–5. Lind PA, Marks LB, Hollis D, et al. Receiver operating characteristic curves to assess predictors of radiation-induced symptomatic lung injury. Int J Radiat Oncol Biol Phys 2002;54:340–7. Graham MV, Purdy JA, Emami B, et al. Clinical dose-volume histogram analysis for pneumonitis after 3D treatment for non-small cell lung cancer (NSCLC). Int J Radiat Oncol Biol Phys 1999;45:323–9. Lonski P, Keehan S, Siva S, et al. Out-of-field in vivo dosimetry using TLD in SABR for primary kidney cancer involving mixed photon fields. Phys Med 2017;37:9–15. Katsuta Y, Kadoya N, Fujita Y, et al. Clinical impact of dosimetric changes for volumetric modulated arc therapy in log file-based patient dose calculations. Phys Med 2017;42:1–6. Katsuta Y, Kadoya N, Fujita Y, et al. Quantification of residual dose estimation error on log file-based patient dose calculation. Phys Med 2016;32:701–5.
51
Contents lists available at ScienceDirect
Physica Medica journal homepage: www.elsevier.com/locate/ejmp
Original paper
Evaluation of functionally weighted dose-volume parameters for thoracic stereotactic ablative radiotherapy (SABR) using CT ventilation
T
⁎
Takayuki Kanaia,b, Noriyuki Kadoyaa, , Yujiro Nakajimaa,c, Yuya Miyasakaa,b, Yoshiro Iekoa,b, Tomohiro Kajikawaa, Kengo Itoa, Takaya Yamamotoa, Suguru Dobashid, Ken Takedad, Kenji Nemotob, Keiichi Jingua a
Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan Department of Radiation Oncology, Yamagata University Faculty of Medicine, Yamagata, Japan c Tokyo Metropolitan Cancer and Infectious Diseases Center Komagome Hospital, Tokyo, Japan d Department of Radiological Technology, School of Health Sciences, Faculty of Medicine, Tohoku University, Sendai, Japan b
A R T I C LE I N FO
A B S T R A C T
Keywords: Functional imaging CT ventilation Radiation pneumonitis Functional planning Thoracic cancer
For the purpose of reducing radiation pneumontisis (RP), four-dimensional CT (4DCT)-based ventilation can be used to reduce functionally weighted lung dose. This study aimed to evaluate the functionally weighted dosevolume parameters and to investigate an optimal weighting method to realize effective planning optimization in thoracic stereotactic ablative radiotherapy (SABR). Forty patients treated with SABR were analyzed. Ventilation images were obtained from 4DCT using deformable registration and Hounsfield unit-based calculation. Functionally-weighted mean lung dose (fMLD) and functional lung fraction receiving at least x Gy (fVx) were calculated by two weighting methods: thresholding and linear weighting. Various ventilation thresholds (5th-95th, every 5th percentile) were tested. The predictive accuracy for CTCAE grade ≧ 2 pneumonitis was evaluated by area under the curve (AUC) of receiver operating characteristic analysis. AUC values varied from 0.459 to 0.570 in accordance with threshold and dose-volume metrics. A combination of 25th percentile threshold and fV30 showed the best result (AUC: 0.570). AUC values with fMLD, fV10, fV20, and fV40 were 0.541, 0.487, 0.548 and 0.563 using a 25th percentile threshold. Although conventional MLD, V10, V20, V30 and V40 showed lower AUC values (0.516, 0.477, 0.534, 0.552 and 0.527), the differences were not statistically significant. fV30 with 25th percentile threshold was the best predictor of RP. Our results suggested that the appropriate weighting should be used for better treatment outcomes in thoracic SABR.
1. Introduction CT ventilation is a state-of-the-art imaging modality using four-dimensional computed tomography (4DCT) and deformable image registration [1]. Its clinical validation studies have been done by comparing other ventilation modalities, such as nuclear medicine [2–6], Xenon-CT [7], hyperpolarized 3He-MRI [8], and pulmonary function test [3,9]. These studies showed reasonable correlations between CT ventilation and other modalities. Because 4DCT has already been used in clinical routine and high resolution ventilation maps can easily be calculated only by image processing, this technique is more suitable for high-precision radiotherapy than the other ventilation modalities. These advantages have strongly motivated currently undergoing clinical trials (NCT02843568, NCT02308709, NCT02528942). These trials ⁎
have been aimed for preservation of post-treatment lung function [10,11] and reduction of radiation pneumonitis (RP) [12–15] by avoiding irradiation to highly functioning lung region [16]. However, different optimization methods were used in each research group. There are two major types of reported optimization strategy to consider inhomogeneity of lung function. The first utilizes the threshold to divide functioning and non-functioning area. Yamamoto et al. [12] used 33th and 66th percentile to divide the total lung into three equal volumes: high, moderate, and low-functional lung. They set optimization constraints in each region. Kadoya et al. [15] used 90th percentile for thresholding the functional lung for planning optimization. The second one is linear weighting. Yamamoto et al. [17] and Kida et al. [18] converted ventilation values into percentile and set weight values linearly to the percentile ventilation. Functionally
Corresponding author at: Department of Radiation Oncology, Tohoku University Graduate School of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai 980-8574, Japan. E-mail address: [email protected] (N. Kadoya).
https://doi.org/10.1016/j.ejmp.2018.05.001 Received 23 March 2018; Received in revised form 27 April 2018; Accepted 1 May 2018 Available online 08 May 2018 1120-1797/ © 2018 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd. All rights reserved.
Physica Medica 49 (2018) 47–51
T. Kanai et al.
Deformable image registration was applied between maximum-inhale CT image HUin(x) and the maximum-exhale image HUex(x). We then calculated the deformation vector u(x) and deformed inhale image HUdeformed(x + u(x)) in each voxel x. The registration process was performed by the elastix [23]: open source registration toolkit. We used a previously reported parameter setting (parameter 2 in [24]) because the mean registration error for lung tissue was reasonably small (< 1.3 mm). All registration results were visually inspected to assure registration accuracy. A Gaussian filter smoothing with a variance σ2 = 1.5 mm2 was subsequently applied to reduce the influence of CT noise [5]. Ventilation value was calculated by Hounsfield unit-based metric (VHU) [1] as:
weighted dose distribution was calculated by the product of weight value and physical dose and used for input to an inverse optimization process. The latter strategy has also been used to correlate functionally weighted dose-volume parameters with incidence of pneumonitis [19]. The optimal weighting strategy which should be used for clinical implementation has been studied by Faught et al. [20] They studied 70 patients treated with conventionally fractionated radiotherapy (CFRT) and concluded that the thresholding-based approach with 84th percentile was the most correlated to grade 3 and higher RP. However, there are no studies which report the correlation between RP and functional dose-volume metrics in thoracic stereotactic ablative radiotherapy (SABR). SABR is characterized as highly conformal dose distribution and hypo-fractionation. These characteristics are known to affect the dose–response of RP[21]. To the best of our knowledge, this is the first study which attempts to quantify clinical significance of CT ventilation in a SABR cohort. We performed receiver operating characteristics (ROC) analysis and evaluated a predictive accuracy of RP by area under the curve (AUC). We also investigated the dependency of AUC values with different weighting approach; thresholding and linear-weighting.
VHU (x) = 1000
W(x,v) =
2.1. Patients and images
The predictive accuracy of grade 2 or greater pneumonitis (CTCAE version 4.0) following radiotherapy was assessed by AUC value of ROC analysis. AUC value is a quantitative value which can assess predictive accuracy of particular indices. AUC value has been used in many studies [19,25,26] because the different prediction from different dose-volume parameters can be easily understood. ROC analysis was performed by in-house MATLAB (The MathWorks Inc., Massachusetts, USA) software. This analysis enabled us to quantify how well different kinds of dosevolume parameters can predict the incidence of radiation pneumonitis. Additionally, the correlation between dose-volume parameters and pneumonitis was also evaluated by univariate analysis. Equality of variance and normality had been confirmed by F-test and Shapiro-Wilk test. Student’s t test was then applied to assess statistical significance with a significance level of 0.05. Statistical analysis was performed by JMP® Pro ver. 12.2.0 (SAS Institute Inc., North Carolina, USA).
For a correction of different fractionation schemes, physical dose distributions were firstly converted into biologically equivalent dose in 2 Gy/fraction (EQD2) using linear-quadratic (LQ) model-based calculation with α / β = 3 [22]. Table 1 Patients’ characteristics and incidence of radiation pneumonitis (RP).
32 8
RP Grade 0 Grade 1 Grade 2
18 13 9
Dose prescription (Gy) Total dose Dose per fraction
40–72 (median 40) 3–10 (median 10)
(2)
2.3. ROC analysis and statistical analysis
2.2. Calculation of functionally weighted dose-volume parameters
58–89 (median 77)
v 50
Above equation means that the highest functioning lung is twice as sensitive to radiation pneumonitis. We obtained weighted dose distribution in each voxel by the product of weight function and dose. Functionally-weighted mean lung dose (fMLD) and functional lung fraction receiving at least 10, 20, 30 and 40 Gy (fV10, fV20, fV30 and fV40) were calculated by the weighted dose distribution. In thresholding method, 19 different percentile thresholds (from 5th to 95th, every 5th percentile) were tested. Functional dose-volume metrics (fMLD, fV10, fV20, fV30 and fV40) were calculated only within voxels which have higher ventilation values than those thresholds.
This study was a retrospective single-institution analysis approved by our institutional review board. Between February 2013 and August 2016, all patients who underwent a pre-treatment 4DCT scan and subsequent thoracic SABR were included. Patients with the following conditions were excluded: (1) History of thoracic radiotherapy, (2) Concurrent chemotherapy, and (3) Follow-up period less than 90 days. In total, 40 patients were analyzed. Patients’ characteristics are summarized in Table 1. Incidence of pneumonitis was evaluated by Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. 4DCT images were acquired as part of clinical therapy to determine internal margin using GE Light Speed RT16 (GE Healthcare, Wisconsin, USA) and Real-time Position Management System (Varian Medical Systems, California, USA). X ray tube current and voltage were 120 mA and 120 kV, respectively. The matrix size in the axial plane was 512 × 512. The axial field of view was determined to include patients’ body. Slice thickness was 2.5 mm. 4DCT images were reconstructed in ten respiratory phases using Advantage 4D workstation (GE Healthcare, Wisconsin, USA)
Age (y)
(1)
To follow other previous studies [12,15,18,19], every ventilation value was converted into percentile. The lowest and the highest ventilation value corresponded to 0th percentile and 100th percentile, respectively. We tested linear weighting method and thresholding method. In linear weighting method, the weight function W(x,v) in each voxel (x) was calculated by percentile ventilation (v) as:
2. Materials and methods
Gender Male Female
HUdeformed ( x+ u(x))−HUex (x) +1 HUex (x){HUdeformed ( x+ u(x)) + 1000}
3. Results Median follow-up period was 14.8 months (range: 3.5–36.9 months). Grade 0, 1, and 2 pneumonitis were observed in 18, 13, and 9 patients, respectively. Fig. 1 shows the summary of AUC values with different methods; thresholding method, linear weighting method and conventional dose-volume parameters. Higher AUC value indicates higher predictive accuracy of the dose-volume parameter. A combination of fV30 and 25th percentile threshold showed the highest predictive accuracy (AUC: 0.570). Twenty-fifth percentile threshold also resulted in good AUC value with fMLD (0.541), fV20 (0.548) and fV40 (0.563), while lower AUC value with fV10 (0.487). These AUC values were higher than those with conventional dose-volume 48
Physica Medica 49 (2018) 47–51
T. Kanai et al.
Fig. 1. Summary of AUC values.
percentile threshold and fV10 with 5th percentile threshold). The difference of AUC values was 0.11. In other words, risk of pneumonitis in 11% of patients can be misestimated by inappropriate choice of dosevolume parameters and weighting method. Therefore, appropriate weighting technique should be used in functional radiotherapy planning. It should be noted that the difference did not reach statistically significant level (p = 0.24).
parameters (0.516, 0.477, 0.534, 0.552 and 0.527 with MLD, V10, V20, V30 and V40, respectively). However, the differences of AUC values were small (ΔAUC < 0.04). Most of the functional dose-volume parameters with linear weighting method showed lower AUC values (0.484, 0.516, 0.520, 0.538 and 0.534 with fMLD, fV10, fV20, fV30 and fV40) than those with conventional dose-volume parameters. The lowest AUC value (0.459) was found in a combination of fV10 and 5th percentile threshold. Any functional or conventional dose-volume parameters did not reach statistically significant level (p > 0.28). Dose distribution and percentile ventilation maps for two representative cases are shown in Fig. 2. In this figure, patient 1 had slightly higher dose-volume parameters (MLD = 5.32 Gy) than patient 2 (MLD = 4.65 Gy). However, high dose areas corresponded to low ventilated region in patient 1 and high ventilated region in patient 2. As a result, functionally-weighted dose-volume parameters with 25th percentile threshold decreased in patient 1 (fMLD = 4.16 Gy) who was not suffered from pneumonitis, and increased in patient 2 (fMLD = 5.29 Gy) who developed grade 2 pneumonitis. Fig. 3 shows the best and the worst ROC curves (fV30 with 25th
4. Discussion We evaluated the predictive accuracy of various functional or conventional dose-volume parameters in SABR. According to Fig. 1, twentieth to thirtieth percentile threshold showed high AUC values, while 70th–80th showed low AUC values. On the contrast, Faught et al. [20] reported 84th percentile resulted in the highest AUC value in CFRT. As compared to CFRT, the lower ventilated lung may contribute the incidence of radiation pneumonitis in hypo-fractionation scheme. In many cases of our results, V30 was more predictive than V20, which commonly used in conventionally fractionated radiotherapy
Fig. 2. Dose distribution and percentile ventilation maps for example cases. 49
Physica Medica 49 (2018) 47–51
T. Kanai et al.
parameters varied with weighting techniques, functional thresholds and dosimetric thresholds. fV30 with 25th percentile would be a good predictor of RP. Our results are useful for functionally optimized planning in SABR. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. References [1] Guerrero T, Sanders K, Castillo E, et al. Dynamic ventilation imaging from fourdimensional computed tomography. Phys Med Biol 2006;51:777–91. [2] Castillo R, Castillo E, McCurdy M, et al. Spatial correspondence of 4D CT ventilation and SPECT pulmonary perfusion defects in patients with malignant airway stenosis. Phys Med Biol 2012;57:1855–71. [3] Yamamoto T, Kabus S, Lorenz C, et al. Pulmonary ventilation imaging based on 4dimensional computed tomography: comparison with pulmonary function tests and SPECT ventilation images. Int J Radiat Oncol Biol Phys 2014;90:414–22. [4] Vinogradskiy Y, Koo PJ, Castillo R, et al. Comparison of 4-dimensional computed tomography ventilation with nuclear medicine ventilation-perfusion imaging: a clinical validation study. Int J Radiat Oncol Biol Phys 2014;89:199–205. [5] Kanai T, Kadoya N, Ito K, et al. Evaluation of four-dimensional computed tomography (4D-CT)-based pulmonary ventilation: the high correlation between 4D-CT ventilation and (81m)Kr-planar images was found. Radiother Oncol 2016;119:444–8. [6] Kipritidis J, Siva S, Hofman MS, et al. Validating and improving CT ventilation imaging by correlating with ventilation 4D-PET/CT using 68Ga-labeled nanoparticles. Med Phys 2014;41:011910. [7] Zhang GG, Latifi K, Du K, et al. Evaluation of the DeltaV 4D CT ventilation calculation method using in vivo xenon CT ventilation data and comparison to other methods. J Appl Clin Med Phys 2016;17:550–60. [8] Mathew L, Wheatley A, Castillo R, et al. Hyperpolarized (3)He magnetic resonance imaging: comparison with four-dimensional x-ray computed tomography imaging in lung cancer. Acad Radiol 2012;19:1546–53. [9] Brennan D, Schubert L, Diot Q, et al. Clinical validation of 4-dimensional computed tomography ventilation with pulmonary function test data. Int J Radiat Oncol Biol Phys 2015;92:423–9. [10] Shioyama Y, Jang SY, Liu HH, et al. Preserving functional lung using perfusion imaging and intensity-modulated radiation therapy for advanced-stage non-small cell lung cancer. Int J Radiat Oncol Biol Phys 2007;68:1349–58. [11] Ohashi T, Takeda A, Shigematsu N, et al. Differences in pulmonary function before vs. 1 year after hypofractionated stereotactic radiotherapy for small peripheral lung tumors. Int J Radiat Oncol Biol Phys 2005;62:1003–8. [12] Yamamoto T, Kabus S, von Berg J, et al. Impact of four-dimensional computed tomography pulmonary ventilation imaging-based functional avoidance for lung cancer radiotherapy. Int J Radiat Oncol Biol Phys 2011;79:279–88. [13] Faught AM, Miyasaka Y, Kadoya N, et al. Evaluating the toxicity reduction with computed tomographic ventilation functional avoidance radiation therapy. Int J Radiat Oncol Biol Phys 2017;99:325–33. [14] Waxweiler T, Schubert L, Diot Q, et al. A complete 4DCT-ventilation functional avoidance virtual trial: developing strategies for prospective clinical trials. J Appl Clin Med Phys 2017. [15] Kadoya N, Cho SY, Kanai T, et al. Dosimetric impact of 4-dimensional computed tomography ventilation imaging-based functional treatment planning for stereotactic body radiation therapy with 3-dimensional conformal radiation therapy. Pract Radiat Oncol 2015;5:e505–12. [16] Doi Y, Kimura T, Nakashima T, et al. Functional image-guided radiotherapy planning in volumetric modulated arc therapy for patients with malignant pleural mesothelioma. Adv Radiat Oncol 2017. [17] Yamamoto T, Kabus S, Bal M, et al. The first patient treatment of computed tomography ventilation functional image-guided radiotherapy for lung cancer. Radiother Oncol 2016;118:227–31. [18] Kida S, Bal M, Kabus S, et al. CT ventilation functional image-based IMRT treatment plans are comparable to SPECT ventilation functional image-based plans. Radiother Oncol 2016;118:521–7. [19] Vinogradskiy Y, Castillo R, Castillo E, et al. Use of 4-dimensional computed tomography-based ventilation imaging to correlate lung dose and function with clinical outcomes. Int J Radiat Oncol Biol Phys 2013;86:366–71. [20] Faught AM, Yamamoto T, Castillo R, et al. Evaluating which dose-function metrics are most critical for functional-guided radiation therapy. Int J Radiat Oncol Biol Phys 2017;99:202–9. [21] Borst GR, Ishikawa M, Nijkamp J, et al. Radiation pneumonitis in patients treated for malignant pulmonary lesions with hypofractionated radiation therapy. Radiother Oncol 2009;91:307–13. [22] Borst GR, Ishikawa M, Nijkamp J, et al. Radiation pneumonitis after hypofractionated radiotherapy: evaluation of the LQ(L) model and different dose parameters. Int J Radiat Oncol Biol Phys 2010;77:1596–603. [23] Klein S, Staring M, Murphy K, et al. Elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging 2010;29:196–205. [24] Kanai T, Kadoya N, Ito K, et al. Evaluation of accuracy of B-spline transformation-
Fig. 3. Examples of ROC curves.
[25,27]. Borst et al. [22] studied radiation pneumonitis in hypo-fractionated radiotherapy and reported the best fitting of normal tissue complication probability curve using high dosimetric thresholds, such as 50 Gy. Our results also suggested that higher dosimetric threshold than 20 Gy would be appropriate for prediction of radiation pneumonitis in SABR. Some clinical trials using CT ventilation (NCT02308709, NCT02843568) includes SABR patients as well as CFRT. However, differences between SABR and CFRT in functional planning have yet to be investigated. Our results suggested that the optimal functional dosevolume parameters in SABR may be different from those of CFRT. Although there is a limitation of small sample size, this study is informative for further clinical trials using CT ventilation in a SABR cohort. In this study, physical dose distributions were calculated by treatment planning systems used in our clinical routine. However, the simulated dose distributions in clinically used treatment planning systems are inaccurate in out-of-field area [28] and does not consider the mis-calibration of leaf, gantry, and collimator [29,30]. It should be noted that we ignored these minor errors between simulated and actual dose distributions. CT ventilation images can be affected by the error in deformable image registration [31,32]. Therefore, we used the previously verified registration algorithm and the parameter setting [24]. The registration algorithm and parameter setting are important factors for the accuracy of deformable image registration. Further improvement could be possible by improved registration algorithms [33,34] or adjusting the parameter setting [35]. Biological property of hypo-fractionated radiotherapy is fundamentally different from CFRT. We used LQ model-based calculation to combine different fractionations because in vivo clinical studies showed LQ model is valid to calculate biological equivalent dose even for hypofractionated radiotherapy [22]. However, some cell experiments have suggested inaccuracy of LQ model in high dose per fraction [36]. Some more complex but sophisticated models, such as modified LQ model [37] or lethal–potentially lethal model [38], may be able to improve prediction for radiation pneumonitis in hypo-fractionated irradiation. 5. Conclusion Predictive
accuracy
of
functionally
weighted
dose-volume 50
Physica Medica 49 (2018) 47–51
T. Kanai et al.
[25]
[26]
[27]
[28] [29]
[30]
[31] Castillo E, Castillo R, Vinogradskiy Y, et al. The numerical stability of transformation-based CT ventilation. Int J Comput Assist Radiol Surg 2017. [32] Yamamoto T, Kabus S, Klinder T, et al. Four-dimensional computed tomography pulmonary ventilation images vary with deformable image registration algorithms and metrics. Med Phys 2011;38:1348–58. [33] Samavati N, Velec M, Brock K. A hybrid biomechanical intensity based deformable image registration of lung 4DCT. Phys Med Biol 2015;60:3359–73. [34] Sarrut D, Baudier T, Ayadi M, et al. Deformable image registration applied to lung SBRT: usefulness and limitations. Phys Med 2017;44:108–12. [35] Miura H, Ozawa S, Nakao M, et al. Impact of deformable image registration accuracy on thoracic images with different regularization weight parameter settings. Phys Med 2017;42:108–11. [36] Puck TT, Morkovin D, Marcus PI, et al. Action of x-rays on mammalian cells. II. Survival curves of cells from normal human tissues. J Exp Med 1957;106:485–500. [37] Guerrero M, Li XA. Extending the linear-quadratic model for large fraction doses pertinent to stereotactic radiotherapy. Phys Med Biol 2004;49:4825–35. [38] Curtis SB. Lethal and potentially lethal lesions induced by radiation–a unified repair model. Radiat Res 1986;106:252–70.
based deformable image registration with different parameter settings for thoracic images. J Radiat Res 2014;55:1163–70. Tsujino K, Hirota S, Endo M, et al. Predictive value of dose-volume histogram parameters for predicting radiation pneumonitis after concurrent chemoradiation for lung cancer. Int J Radiat Oncol Biol Phys 2003;55:110–5. Lind PA, Marks LB, Hollis D, et al. Receiver operating characteristic curves to assess predictors of radiation-induced symptomatic lung injury. Int J Radiat Oncol Biol Phys 2002;54:340–7. Graham MV, Purdy JA, Emami B, et al. Clinical dose-volume histogram analysis for pneumonitis after 3D treatment for non-small cell lung cancer (NSCLC). Int J Radiat Oncol Biol Phys 1999;45:323–9. Lonski P, Keehan S, Siva S, et al. Out-of-field in vivo dosimetry using TLD in SABR for primary kidney cancer involving mixed photon fields. Phys Med 2017;37:9–15. Katsuta Y, Kadoya N, Fujita Y, et al. Clinical impact of dosimetric changes for volumetric modulated arc therapy in log file-based patient dose calculations. Phys Med 2017;42:1–6. Katsuta Y, Kadoya N, Fujita Y, et al. Quantification of residual dose estimation error on log file-based patient dose calculation. Phys Med 2016;32:701–5.
51