- Email: [email protected]
Multi-staged robotic stereotactic radiosurgery for large cerebral arteriovenous malformations
Radiotherapy and Oncology 109 (2013) 452–456
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Stereotactic radiotherapy
Multi-staged robotic stereotactic radiosurgery for large cerebral arteriovenous malformations Chuxiong Ding a,⇑, Timothy D. Solberg a, Brian Hrycushko a, Paul Medin a, Louis Whitworth a,b, Robert D. Timmerman a a
Department of Radiation Oncology; b Department of Neurological Surgery, University of Texas Southwestern Medical Center, Dallas, USA
a r t i c l e
i n f o
Article history: Received 19 February 2013 Received in revised form 18 July 2013 Accepted 24 July 2013 Available online 7 September 2013 Keywords: CyberKnife Stereotactic radiosurgery Arteriovenous malformations
a b s t r a c t Purpose: To investigate a multi-staged robotic stereotactic radiosurgery (SRS) delivery technique for the treatment of large cerebral arteriovenous malformations (AVMs). The treatment planning process and strategies to optimize both individual and composite dosimetry are discussed. Methods: Eleven patients with large (30.7 ± 19.2 cm3) AVMs were selected for this study. A fiducial system was designed for fusion of targets between planar angiograms and simulation CT scans. AVMs were contoured based on single contrast CT, MRI and orthogonal angiogram images. AVMs were divided into 3–8 sub-target volumes (3–7 cm3) for sequential treatment at 1–4 week intervals to a prescription dose of 16–20 Gy. Forward and inversely developed treatment plans were optimized for 95% coverage of the total AVM volume by dose summation from each sub-volume, while minimizing dose to surrounding tissues. Dose-volume analysis was used to evaluate the PTV coverage, dose conformality (CI), and R50 and V12Gy parameters. Results: The treatment workflow was commissioned and able to localize within 1 mm. Inverse optimization outperformed forward planning for most patients for each index considered. Dose conformality was shown comparable to staged Gamma Knife treatments. Conclusion: The CyberKnife system is shown to be a practical delivery platform for multi-staged treatments of large AVMs using forward or inverse planning techniques. Ó 2013 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 109 (2013) 452–456
Radiosurgery has been established as an effective treatment for small cerebral arteriovenous malformations (AVM) [1]. Radiation is believed to induce endothelial cell proliferation leading to luminal closure, and total obliteration of the AVM nidus is achieved in 70– 95% of patients at 3–5 years [2–4]. Obliteration of larger AVMs (>10–15 cm3 or diameter P3 cm) has been less successful using radiosurgery [5] and other treatment modalities [6]. Radiation dose must be decreased with increasing lesion volume to prevent normal tissue toxicity, consequently limiting rates of obliteration [7]. In-field obliteration is well-correlated with the minimum radiosurgical dose [2], thus several centers have pursued a multistaged, spatially-fractionated treatment of large AVMs using a Gamma Knife treatment platform (Elekta, Stockholm, Sweden) [8,9]. Staging AVM treatments by segmenting AVMs into sub-volumes treated as separate targets over a period of several months permits a more potent dose than could be delivered to the entire volume in a single fraction. With this approach, it is hypothesized ⇑ Corresponding author. Address: Department of Radiation Oncology, University of Texas Southwestern Medical Center, 5801 Forest Park Road, Dallas, TX 75390, USA. E-mail address: [email protected] (C. Ding). 0167-8140/$ - see front matter Ó 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.radonc.2013.07.018
that increased radiation dose reduces intracranial hemorrhage risks for large AVMs by improving obliteration rates, and normal tissues surrounding the nidus undergo sub-lethal radiation repair prior to the next stage. CyberKnife (Accuray, Sunnyvale, CA) is an image-guided SRS system [10] that can deliver multiple isocentric or non-isocentric beams to a desired target using a 6 MV linac mounted on a robotic arm. An integrated stereoscopic kilovoltage (kV) imaging system localizes and monitors the patient position throughout a course of treatment by matching bony anatomy to digitally reconstructed radiographs (DRRs) generated from CT simulation. The CyberKnife platform has been shown to be a safe and efficient delivery platform for single-fraction radiosurgery treatment of small AVMs [11], and would be beneficial in delivering multi-staged treatments for large AVMs. The CyberKnife MultiPlan treatment planning system provides for a composite dose distribution of all stages prior to treatment, and patients can be easily re-positioned for multiple frameless fractions with kV imaging. In this work, we present a practical methodology and clinical protocol for multi-staged robotic SRS of large cerebral AVMs using forward and inverse treatment planning techniques.
C. Ding et al. / Radiotherapy and Oncology 109 (2013) 452–456
Materials and methods Patient selection This study was approved by the institutional review board and conducted in accordance with institutional guidelines. Eleven patients diagnosed with large (30.7 ± 19.2 cm3) localized cerebral AVMs were selected for multi-stage robotic SRS. Patient imaging Patients were immobilized and imaged using a 16-slice Brilliance Big Bore CT scanner (Philips Healthcare, Andover, MD) without contrast to be used as the primary image set for treatment planning and for future patient setup with the CyberKnife delivery system. A custom-built angiogram localization frame (Fig. 1(a)) was attached to the head immobilization system and a contrast CT scan was acquired with the patient in the identical position. A T1-weighted MRI image was acquired with gadolinium contrast for improved target definition. Fiducial markers of the custom localization frame provide coordinates and a reference geometry for bi-planar angiography. Following image acquisition, a radiologist manually delineated the AVM nidus on the angiogram images. The nidus of the AVM was registered to the CT image set following digitization of the 2D angiogram films.
453
Target localization The non-contrast CT image set was registered with the contrast CT set using the DICOM coordinates. The non-contrast CT and MRI were rigidly registered within the MultiPlan treatment planning system. All registered images were transferred to the Pinnacle3 treatment planning system (Phillips Medical Systems, Cleveland, OH) for contouring. AP and lateral angiogram films (Fig. 1(b and c)) were placed on a digitizer (A56BL.H) (Numonics Corporation, San Diego, CA) to outline the nidus position relative to the CT images. As shown in Fig. 1(d), two beams, which have the identical source position and beam projection direction as the angiogram images, were added in the Pinnacle treatment planning system to be used as a reference geometry. The source to isocenter distance and magnification factor were obtained using the following equations:
SADPA ¼
fPA ¼
DAP ðdSIAP þ dLR Þ 2ðdSIAP dLR Þ
ð1Þ
ðSADPA DAP =2ÞdSI:AP Df SFDPA
ð2Þ
DLR ðdSI:LR þ dAP Þ 2ðdSI:LR dAP Þ
ð3Þ
SADLR ¼
Fig. 1. Frame used for angiographic target localization (a), target localization on patient simulation CT using angiogram frame (b), target region delineated on digitized angiogram to be projected on DRRs in treatment planning system (c), and geometry of angiographic frame system in treatment planning system (d).
454
fLR ¼
Multi-stage robotic stereotactic radiosurgery for large AVMs
ðSADLR DLR =2ÞdSI:LR Df SFDLR
ð4Þ
As shown in Fig. 1 (b and c), DAP (305 mm) and DLR (260 mm) are defined by the localization frame geometry; Df (80 mm) is the distance between the first and last fiducial markers on each side of the frame; dSI.AP, dLR, dSI.LR, and dAP are the distances between the distal fiducial markers on film. SADPA and SADLR are the source to isocenter distance for PA and lateral beams, respectively; fPA and fLR are the magnification factors for the PA and lateral DRRs to the corresponding angiogram films, respectively; and SFDPA and SFDLR are the source to film distances for the PA and lateral DRRs in the Pinnacle treatment planning system. The block digitizing tool in Pinnacle was used to create blocks (Fig. 1(d)) for each beam based on contours of the AVM from AP and lateral angiograms. A target volume was obtained by contouring the overlapped blocked projection areas from both AP and lateral beams. Additional information from the contrast CT and MRI datasets was used to contour the final AVM nidus. Depending on the target size and shape, each AVM was divided manually into 3–8 sub-targets with approximately equal volumes (6.9 ± 3.2 cm3) as shown in Fig. 2(a). Treatment planning Forward planning technique CyberKnife plans were generated for each sub-target using cone sizes ranging from 60–80% of the minimum dimension of the subtarget. An iterative forward planning technique was used to create treatment plans with greater than or equal to 95% coverage of the total AVM volume following summation of the dose from each of the sub-volumes, while minimizing the dose to the surrounding normal tissues. A volume (30–60% of the sub-target volume) was created within each sub-target for beam targeting to be fully covered by the prescription dose (as shown in Fig. 2(a)). Each sub-target dose was summed as shown in Fig. 2(a) and dose-volume histogram (DVH) and isodose distributions were then analyzed. Modification of the composite dose distribution was repeated to optimize the plan until the dose constraints for the AVM and surrounding tissues were met. Inverse planning technique A novel inverse planning technique was developed to optimize the dose distribution on the target while sparing normal tissue. The Multiplan inverse planning technique is used to develop an optimal plan for the entire AVM target that satisfies the prescribed objectives. Using each beam’s source position, the target position and monitor units (MUs), a line in space is generated for each beam axis from the source to target using the following equation.
x xs y ys z zs ¼ ¼ xs xt ys yt zs zt
ð5Þ
where x, y, and z are coordinates of the point on the line in space, xs, ys, and zs are coordinates of the CyberKnife system source and xt, yt, and zt are coordinates of the target. In-house software takes the contours of the sub-targets of the AVM from DICOM files exported by the Multiplan treatment planning system and calculates the path-length of each beam through individual sub-targets. The total MUs of each beam, as calculated by inverse planning on the total AVM target, are reduced by weighting the ratio of the path length of the beam through a sub-target to the total path length of all the beams through the sub-target. This procedure is illustrated in Fig. 2(b). For the i-th beam, with MUi, passing through sub-targets 1 and 2, the path-length in the j-th sub-target is dij. Therefore, the MUij for the i-th beam in the stage j, which is planned for subtarget j is obtained using the following equation:
MU ij ¼ MU i
dij N X dik
ð6Þ
k¼1
where N is the total number of sub-targets. Each of the sub-targets is treated in a different stage with the same beams used in the original inverse plan, but with the MUs altered depending on the beams’ importance in delivering dose to an individual sub-target. The composite dose of the staged plan is identical to the original plan which was inversely optimized to the entire AVM target. Dose volume characteristics For this study, the composite prescription dose covered 95% of the total AVM volume. The dose conformality index recommended by the RTOG for radiosurgery (PITV; ratio of total volume receiving the prescription dose or greater to that of the target volume), the Paddick conformity index (CI) [12], the R50 (ratio of the volume receiving 50% of the prescription dose or greater to the target volume), and the V12Gy (ratio of volume receiving 12 Gy or greater to the volume of the whole brain) indices were used to evaluate the final PTV and normal tissue dose coverage. For planning purposes the PITV was kept as low as achievable and constrained to be less than 2.0. The prescription dose ranged from 16 Gy to 20 Gy depending on the volume of the AVM target.
Quality assurance An anthropomorphic head phantom, shown in Fig. 1(a), was used to assess the localization accuracy of this system. A CT scan of an anthropomorphic head phantom was acquired with the localization frame attached. Bi-planar radiographic images of the phantom were obtained with the central fiducial markers of the frame aligned together in the AP and lateral directions. Fiducials inside the phantom were contoured on AP and lateral films and digitized into the Pinnacle treatment planning system. A passing criterion of 61 mm was set for the alignment of fiducial centers between the digitized film contour and the CT image set. Individual patient QA was performed before each treatment. Three gold fiducial markers were placed on the patient’s mask and were contoured on AP and lateral angiogram films to be digitized into the Pinnacle treatment planning system. The passing criterion for this test was that the alignment of the fiducial centers between the digitized film contour and the CT image set be within 1 mm. Results As discussed in the methods section, the entire treatment workflow was successfully commissioned for the treatment of large AVMs using the CyberKnife delivery and treatment planning systems. Based on quality assurance tests, stereotactic localization of the AVM target is achievable within 1 mm. Table 1 shows the parameters used for planning and evaluation for the eleven patients in this study. Dose distributions were optimized to achieve approximately 95% PTV coverage with a PITV conformity index as low as achievable and constrained to be less than 2.0. Eight of the eleven patients had better target coverage by the prescription dose for inversely optimized plans (96.0 ± 0.6%) compared with the forward optimized plans (95.8 ± 1.3%) while all of the patients had improved high dose conformality with the inversely optimized plans (PITV of 1.44 ± 0.16 and 1.23 ± 0.16 for forward and inverse plans, respectively and CI of 0.65 ± 0.06 and 0.76 ± 0.09 for forward and inverse plans, respectively). Moderate dose spillage within the brain, as repre-
C. Ding et al. / Radiotherapy and Oncology 109 (2013) 452–456
455
Fig. 2. Contours and forward planning of sub-targets for a typical patient (a); and the monitor unit splitting scheme for a sub-target plan for the inverse planning procedure (b).
456
Multi-stage robotic stereotactic radiosurgery for large AVMs
Table 1 Summary of dosimetric analysis. Patients 3
AVM (cm ) Treatment stages Prescription dose (Gy) Prescription dose (Gy) coverage (%) CI PITV V12Gy (%) R50
F I F I F I F I F I
1
2
3
4
5
6
7
8
9
10
11
Avg ± SD
67 6 16 94.7 95.7 0.74 0.84 1.21 1.09 7.3 7.2 2.7 2.8
65.8 8 16 94.8 95.8 0.71 0.83 1.27 1.11 9.2 7.3 3.4 2.4
21.3 4 18 94.5 95.6 0.69 0.80 1.29 1.14 3.6 3.2 3.7 3.1
15.3 3 20 95.0 95.5 0.72 0.81 1.25 1.13 2.6 2.3 3.5 3.0
17.7 3 18 95.6 96.2 0.63 0.79 1.44 1.17 3.0 2.4 3.6 3.0
23.9 4 18 95.5 97.1 0.62 0.69 1.47 1.36 4.4 4.0 4.7 4.0
25.4 7 18 97.5 95.6 0.55 0.59 1.73 1.55 4.8 4.7 4.9 4.4
16.3 3 18 95.2 97.1 0.60 0.64 1.52 1.48 2.9 3.1 4.3 4.2
25.1 5 18 95.4 95.8 0.59 0.72 1.54 1.27 5.2 4.2 4.1 3.3
17.4 3 20 97.6 96.4 0.59 0.82 1.62 1.13 4.2 2.3 4.6 2.4
42.7 3 16 98.2 95.5 0.66 0.84 1.46 1.09 6.9 5.1 3.7 2.9
30.7 ± 19.2 NA NA 95.8 ± 1.3 96.0 ± 0.6 0.65 ± 0.06 0.76 ± 0.09
4.9 ± 2.1 4.2 ± 1.8 3.9 ± 0.7 3.2 ± 0.7
PITV: ratio of prescription dose volume to that of target volume; CI: ratio of target volume receiving the prescription isodose to the target volume, multiplied by the ratio of the target volume receiving the prescription isodose to the prescription isodose volume; R50: ratio of volume receiving 50% of the prescription dose to the target volume; V12Gy: percent of brain volume receiving 12 Gy or more; F: forward planned; I: inversely planned.
sented by the R50 parameter, was shown to improve in ten of the eleven patients for inversely optimized plans (3.2 ± 0.7 and 3.9 ± 0.7 for inverse and forward plans, respectively). The V12Gy parameter, used here as a surrogate for normal brain complications, was also shown to improve in ten of the eleven patients for inversely optimized plans (4.2 ± 1.8% and 4.9 ± 2.1% for inverse and forward plans, respectively). Discussion and conclusion The CyberKnife and MultiPlan radiation delivery and treatment planning systems are shown in this work to be practical for multistage SRS of large cerebral AVMs. Highly conformal, composite treatment plans were successfully delivered to 11 patients with similar dose distributions as that of staged Gamma Knife treatments. The CI parameter is a measure of the compactness of the high dose distribution to an intended target as well as the location of the prescription isodose volume with respect to the target volume. A value close to 1.0 is expected to minimize radiation related complications. The average CI for the inverse planning technique in this study was shown to be comparable to that described for single fraction treatment of smaller AVMs using Gamma Knife Perfexion or Cyberknife systems (0.80) [13]. Dose-volume effects and parameters which correlate well with normal tissue complications for the staged treatment of large-volume AVMs have yet to be determined. The V12Gy index has been observed to correlate with the risk of developing SRS-related imaging changes and has been used as a surrogate for normal tissue damage [14]; thus the volume receiving 12 Gy is minimized in staged treatments, regardless of the number of stages used [15]. In this work, the composite V12Gy was less than 10% of the brain volume for all patients. This parameter continues to be used for plan evaluation and comparison as experience is gained in staged treatments of large AVMs. A new inverse optimization planning technique for staged treatments was shown to improve treatment plans in the majority of patients for all indices evaluated compared to forward optimized plans. The increased target conformality and normal tissue sparing allow for the possibility of increasing the prescription dose with the hope of improving nidus obliteration rates. The inverse planning technique developed for this technique is considerably more efficient than forward planning as the latter requires the treatment planner to envision a composite dose distribution while creating the dose distribution for individual sub-targets. This becomes increasingly difficult as the number of sub-targets is increased or for AVMs adjacent to organs at risk, such as the optic pathway.
With further patient follow-up, future work will investigate treatment outcome and toxicity for staged treatments of large AVMs using the CyberKnife delivery platform. Conflict of interest None. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.radonc. 2013.07.018. References [1] Lunsford LD, Niranjan A, Kondziolka D, et al. Arteriovenous malformation radiosurgery: a twenty year perspective. Clin Neurosurg 2008;55:108–19. [2] Flickinger JC, Kondziolka D, Maitz AH, Lunsford LD. An analysis of the doseresponse for arteriovenous malformation radiosurgery and other factors affecting obliteration. Radiother Oncol 2002;63:347–54. [3] Friedman WA, Bova FJ. Radiosurgery for arteriovenous malformations. Neurol Res 2011;33:803–19. [4] Kano H, Kondziolka D, Flickinger JC, et al. Stereotactic radiosurgery for arteriovenous malformations, Part 6: multistaged volumetric management of large arteriovenous malformations. J Neurosurg 2012;116:54–65. [5] Zabel A, Milker-Zabel S, Huber P, et al. Treatment outcome after linac-based radiosurgery in cerebral arteriovenous malformations: retrospective analysis of factors affecting obliteration. Radiother Oncol 2005;77:105–10. [6] Jayaraman MV, Marcellus ML, Do HM, et al. Hemorrhage rate in patients with Spetzler-Martin grades IV and V arteriovenous malformations – is treatment justified? Stroke 2007;38:325–9. [7] Barker FG, Butler WE, Lyons S, et al. Dose-volume prediction of radiationrelated complications after proton beam radiosurgery for cerebral arteriovenous malformations. J Neurosurg 2003;99:254–63. [8] Nagy G, Rowe JG, Radatz MWR, et al. A historical analysis of single-stage gamma knife radiosurgical treatment for large arteriovenous malformations: evolution and outcomes. Acta Neurochir 2012;154:383–94. [9] Back AG, Vollmer D, Zeck O, et al. Retrospective analysis of unstaged and staged Gamma Knife surgery with and without preceding embolization for the treatment of arteriovenous malformations. J Neurosurg 2008;109:57–64. [10] Adler JR, Chang SD, Murphy MJ, et al. The cyberknife: a frameless robotic system for radiosurgery. Stereotact Funct Neurosurg 1997;69:124–8. [11] Hristov D, Liu LN, Adler JR, et al. Technique for targeting arteriovenous malformations using frameless image-guided robotic radiosurgery. Int J Radiat Oncol 2011;79:1232–40. [12] Paddick I. A simple scoring ratio to index the conformity of radiosurgical treatment plans – technical note. J Neurosurg 2000;93:219–22. [13] Gevaert T, Levivier M, Lacornerie T, et al. Dosimetric comparison of different treatment modalities for stereotactic radiosurgery of arteriovenous malformations and acoustic neuromas. Radiother Oncol 2013;106:192–7. [14] Flickinger JC, Kondziolka D, Kalend AM, et al. Radiosurgery-related imaging changes in surrounding brain: multivariate analysis and model evaluation. Radiosurgery 1996;1:229–36. [15] Jones J, Jang S, Getch CC, et al. Advances in the radiosurgical treatment of large inoperable arteriovenous malformations. Neurosurg Focus 2007;23:E7.
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Stereotactic radiotherapy
Multi-staged robotic stereotactic radiosurgery for large cerebral arteriovenous malformations Chuxiong Ding a,⇑, Timothy D. Solberg a, Brian Hrycushko a, Paul Medin a, Louis Whitworth a,b, Robert D. Timmerman a a
Department of Radiation Oncology; b Department of Neurological Surgery, University of Texas Southwestern Medical Center, Dallas, USA
a r t i c l e
i n f o
Article history: Received 19 February 2013 Received in revised form 18 July 2013 Accepted 24 July 2013 Available online 7 September 2013 Keywords: CyberKnife Stereotactic radiosurgery Arteriovenous malformations
a b s t r a c t Purpose: To investigate a multi-staged robotic stereotactic radiosurgery (SRS) delivery technique for the treatment of large cerebral arteriovenous malformations (AVMs). The treatment planning process and strategies to optimize both individual and composite dosimetry are discussed. Methods: Eleven patients with large (30.7 ± 19.2 cm3) AVMs were selected for this study. A fiducial system was designed for fusion of targets between planar angiograms and simulation CT scans. AVMs were contoured based on single contrast CT, MRI and orthogonal angiogram images. AVMs were divided into 3–8 sub-target volumes (3–7 cm3) for sequential treatment at 1–4 week intervals to a prescription dose of 16–20 Gy. Forward and inversely developed treatment plans were optimized for 95% coverage of the total AVM volume by dose summation from each sub-volume, while minimizing dose to surrounding tissues. Dose-volume analysis was used to evaluate the PTV coverage, dose conformality (CI), and R50 and V12Gy parameters. Results: The treatment workflow was commissioned and able to localize within 1 mm. Inverse optimization outperformed forward planning for most patients for each index considered. Dose conformality was shown comparable to staged Gamma Knife treatments. Conclusion: The CyberKnife system is shown to be a practical delivery platform for multi-staged treatments of large AVMs using forward or inverse planning techniques. Ó 2013 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 109 (2013) 452–456
Radiosurgery has been established as an effective treatment for small cerebral arteriovenous malformations (AVM) [1]. Radiation is believed to induce endothelial cell proliferation leading to luminal closure, and total obliteration of the AVM nidus is achieved in 70– 95% of patients at 3–5 years [2–4]. Obliteration of larger AVMs (>10–15 cm3 or diameter P3 cm) has been less successful using radiosurgery [5] and other treatment modalities [6]. Radiation dose must be decreased with increasing lesion volume to prevent normal tissue toxicity, consequently limiting rates of obliteration [7]. In-field obliteration is well-correlated with the minimum radiosurgical dose [2], thus several centers have pursued a multistaged, spatially-fractionated treatment of large AVMs using a Gamma Knife treatment platform (Elekta, Stockholm, Sweden) [8,9]. Staging AVM treatments by segmenting AVMs into sub-volumes treated as separate targets over a period of several months permits a more potent dose than could be delivered to the entire volume in a single fraction. With this approach, it is hypothesized ⇑ Corresponding author. Address: Department of Radiation Oncology, University of Texas Southwestern Medical Center, 5801 Forest Park Road, Dallas, TX 75390, USA. E-mail address: [email protected] (C. Ding). 0167-8140/$ - see front matter Ó 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.radonc.2013.07.018
that increased radiation dose reduces intracranial hemorrhage risks for large AVMs by improving obliteration rates, and normal tissues surrounding the nidus undergo sub-lethal radiation repair prior to the next stage. CyberKnife (Accuray, Sunnyvale, CA) is an image-guided SRS system [10] that can deliver multiple isocentric or non-isocentric beams to a desired target using a 6 MV linac mounted on a robotic arm. An integrated stereoscopic kilovoltage (kV) imaging system localizes and monitors the patient position throughout a course of treatment by matching bony anatomy to digitally reconstructed radiographs (DRRs) generated from CT simulation. The CyberKnife platform has been shown to be a safe and efficient delivery platform for single-fraction radiosurgery treatment of small AVMs [11], and would be beneficial in delivering multi-staged treatments for large AVMs. The CyberKnife MultiPlan treatment planning system provides for a composite dose distribution of all stages prior to treatment, and patients can be easily re-positioned for multiple frameless fractions with kV imaging. In this work, we present a practical methodology and clinical protocol for multi-staged robotic SRS of large cerebral AVMs using forward and inverse treatment planning techniques.
C. Ding et al. / Radiotherapy and Oncology 109 (2013) 452–456
Materials and methods Patient selection This study was approved by the institutional review board and conducted in accordance with institutional guidelines. Eleven patients diagnosed with large (30.7 ± 19.2 cm3) localized cerebral AVMs were selected for multi-stage robotic SRS. Patient imaging Patients were immobilized and imaged using a 16-slice Brilliance Big Bore CT scanner (Philips Healthcare, Andover, MD) without contrast to be used as the primary image set for treatment planning and for future patient setup with the CyberKnife delivery system. A custom-built angiogram localization frame (Fig. 1(a)) was attached to the head immobilization system and a contrast CT scan was acquired with the patient in the identical position. A T1-weighted MRI image was acquired with gadolinium contrast for improved target definition. Fiducial markers of the custom localization frame provide coordinates and a reference geometry for bi-planar angiography. Following image acquisition, a radiologist manually delineated the AVM nidus on the angiogram images. The nidus of the AVM was registered to the CT image set following digitization of the 2D angiogram films.
453
Target localization The non-contrast CT image set was registered with the contrast CT set using the DICOM coordinates. The non-contrast CT and MRI were rigidly registered within the MultiPlan treatment planning system. All registered images were transferred to the Pinnacle3 treatment planning system (Phillips Medical Systems, Cleveland, OH) for contouring. AP and lateral angiogram films (Fig. 1(b and c)) were placed on a digitizer (A56BL.H) (Numonics Corporation, San Diego, CA) to outline the nidus position relative to the CT images. As shown in Fig. 1(d), two beams, which have the identical source position and beam projection direction as the angiogram images, were added in the Pinnacle treatment planning system to be used as a reference geometry. The source to isocenter distance and magnification factor were obtained using the following equations:
SADPA ¼
fPA ¼
DAP ðdSIAP þ dLR Þ 2ðdSIAP dLR Þ
ð1Þ
ðSADPA DAP =2ÞdSI:AP Df SFDPA
ð2Þ
DLR ðdSI:LR þ dAP Þ 2ðdSI:LR dAP Þ
ð3Þ
SADLR ¼
Fig. 1. Frame used for angiographic target localization (a), target localization on patient simulation CT using angiogram frame (b), target region delineated on digitized angiogram to be projected on DRRs in treatment planning system (c), and geometry of angiographic frame system in treatment planning system (d).
454
fLR ¼
Multi-stage robotic stereotactic radiosurgery for large AVMs
ðSADLR DLR =2ÞdSI:LR Df SFDLR
ð4Þ
As shown in Fig. 1 (b and c), DAP (305 mm) and DLR (260 mm) are defined by the localization frame geometry; Df (80 mm) is the distance between the first and last fiducial markers on each side of the frame; dSI.AP, dLR, dSI.LR, and dAP are the distances between the distal fiducial markers on film. SADPA and SADLR are the source to isocenter distance for PA and lateral beams, respectively; fPA and fLR are the magnification factors for the PA and lateral DRRs to the corresponding angiogram films, respectively; and SFDPA and SFDLR are the source to film distances for the PA and lateral DRRs in the Pinnacle treatment planning system. The block digitizing tool in Pinnacle was used to create blocks (Fig. 1(d)) for each beam based on contours of the AVM from AP and lateral angiograms. A target volume was obtained by contouring the overlapped blocked projection areas from both AP and lateral beams. Additional information from the contrast CT and MRI datasets was used to contour the final AVM nidus. Depending on the target size and shape, each AVM was divided manually into 3–8 sub-targets with approximately equal volumes (6.9 ± 3.2 cm3) as shown in Fig. 2(a). Treatment planning Forward planning technique CyberKnife plans were generated for each sub-target using cone sizes ranging from 60–80% of the minimum dimension of the subtarget. An iterative forward planning technique was used to create treatment plans with greater than or equal to 95% coverage of the total AVM volume following summation of the dose from each of the sub-volumes, while minimizing the dose to the surrounding normal tissues. A volume (30–60% of the sub-target volume) was created within each sub-target for beam targeting to be fully covered by the prescription dose (as shown in Fig. 2(a)). Each sub-target dose was summed as shown in Fig. 2(a) and dose-volume histogram (DVH) and isodose distributions were then analyzed. Modification of the composite dose distribution was repeated to optimize the plan until the dose constraints for the AVM and surrounding tissues were met. Inverse planning technique A novel inverse planning technique was developed to optimize the dose distribution on the target while sparing normal tissue. The Multiplan inverse planning technique is used to develop an optimal plan for the entire AVM target that satisfies the prescribed objectives. Using each beam’s source position, the target position and monitor units (MUs), a line in space is generated for each beam axis from the source to target using the following equation.
x xs y ys z zs ¼ ¼ xs xt ys yt zs zt
ð5Þ
where x, y, and z are coordinates of the point on the line in space, xs, ys, and zs are coordinates of the CyberKnife system source and xt, yt, and zt are coordinates of the target. In-house software takes the contours of the sub-targets of the AVM from DICOM files exported by the Multiplan treatment planning system and calculates the path-length of each beam through individual sub-targets. The total MUs of each beam, as calculated by inverse planning on the total AVM target, are reduced by weighting the ratio of the path length of the beam through a sub-target to the total path length of all the beams through the sub-target. This procedure is illustrated in Fig. 2(b). For the i-th beam, with MUi, passing through sub-targets 1 and 2, the path-length in the j-th sub-target is dij. Therefore, the MUij for the i-th beam in the stage j, which is planned for subtarget j is obtained using the following equation:
MU ij ¼ MU i
dij N X dik
ð6Þ
k¼1
where N is the total number of sub-targets. Each of the sub-targets is treated in a different stage with the same beams used in the original inverse plan, but with the MUs altered depending on the beams’ importance in delivering dose to an individual sub-target. The composite dose of the staged plan is identical to the original plan which was inversely optimized to the entire AVM target. Dose volume characteristics For this study, the composite prescription dose covered 95% of the total AVM volume. The dose conformality index recommended by the RTOG for radiosurgery (PITV; ratio of total volume receiving the prescription dose or greater to that of the target volume), the Paddick conformity index (CI) [12], the R50 (ratio of the volume receiving 50% of the prescription dose or greater to the target volume), and the V12Gy (ratio of volume receiving 12 Gy or greater to the volume of the whole brain) indices were used to evaluate the final PTV and normal tissue dose coverage. For planning purposes the PITV was kept as low as achievable and constrained to be less than 2.0. The prescription dose ranged from 16 Gy to 20 Gy depending on the volume of the AVM target.
Quality assurance An anthropomorphic head phantom, shown in Fig. 1(a), was used to assess the localization accuracy of this system. A CT scan of an anthropomorphic head phantom was acquired with the localization frame attached. Bi-planar radiographic images of the phantom were obtained with the central fiducial markers of the frame aligned together in the AP and lateral directions. Fiducials inside the phantom were contoured on AP and lateral films and digitized into the Pinnacle treatment planning system. A passing criterion of 61 mm was set for the alignment of fiducial centers between the digitized film contour and the CT image set. Individual patient QA was performed before each treatment. Three gold fiducial markers were placed on the patient’s mask and were contoured on AP and lateral angiogram films to be digitized into the Pinnacle treatment planning system. The passing criterion for this test was that the alignment of the fiducial centers between the digitized film contour and the CT image set be within 1 mm. Results As discussed in the methods section, the entire treatment workflow was successfully commissioned for the treatment of large AVMs using the CyberKnife delivery and treatment planning systems. Based on quality assurance tests, stereotactic localization of the AVM target is achievable within 1 mm. Table 1 shows the parameters used for planning and evaluation for the eleven patients in this study. Dose distributions were optimized to achieve approximately 95% PTV coverage with a PITV conformity index as low as achievable and constrained to be less than 2.0. Eight of the eleven patients had better target coverage by the prescription dose for inversely optimized plans (96.0 ± 0.6%) compared with the forward optimized plans (95.8 ± 1.3%) while all of the patients had improved high dose conformality with the inversely optimized plans (PITV of 1.44 ± 0.16 and 1.23 ± 0.16 for forward and inverse plans, respectively and CI of 0.65 ± 0.06 and 0.76 ± 0.09 for forward and inverse plans, respectively). Moderate dose spillage within the brain, as repre-
C. Ding et al. / Radiotherapy and Oncology 109 (2013) 452–456
455
Fig. 2. Contours and forward planning of sub-targets for a typical patient (a); and the monitor unit splitting scheme for a sub-target plan for the inverse planning procedure (b).
456
Multi-stage robotic stereotactic radiosurgery for large AVMs
Table 1 Summary of dosimetric analysis. Patients 3
AVM (cm ) Treatment stages Prescription dose (Gy) Prescription dose (Gy) coverage (%) CI PITV V12Gy (%) R50
F I F I F I F I F I
1
2
3
4
5
6
7
8
9
10
11
Avg ± SD
67 6 16 94.7 95.7 0.74 0.84 1.21 1.09 7.3 7.2 2.7 2.8
65.8 8 16 94.8 95.8 0.71 0.83 1.27 1.11 9.2 7.3 3.4 2.4
21.3 4 18 94.5 95.6 0.69 0.80 1.29 1.14 3.6 3.2 3.7 3.1
15.3 3 20 95.0 95.5 0.72 0.81 1.25 1.13 2.6 2.3 3.5 3.0
17.7 3 18 95.6 96.2 0.63 0.79 1.44 1.17 3.0 2.4 3.6 3.0
23.9 4 18 95.5 97.1 0.62 0.69 1.47 1.36 4.4 4.0 4.7 4.0
25.4 7 18 97.5 95.6 0.55 0.59 1.73 1.55 4.8 4.7 4.9 4.4
16.3 3 18 95.2 97.1 0.60 0.64 1.52 1.48 2.9 3.1 4.3 4.2
25.1 5 18 95.4 95.8 0.59 0.72 1.54 1.27 5.2 4.2 4.1 3.3
17.4 3 20 97.6 96.4 0.59 0.82 1.62 1.13 4.2 2.3 4.6 2.4
42.7 3 16 98.2 95.5 0.66 0.84 1.46 1.09 6.9 5.1 3.7 2.9
30.7 ± 19.2 NA NA 95.8 ± 1.3 96.0 ± 0.6 0.65 ± 0.06 0.76 ± 0.09
4.9 ± 2.1 4.2 ± 1.8 3.9 ± 0.7 3.2 ± 0.7
PITV: ratio of prescription dose volume to that of target volume; CI: ratio of target volume receiving the prescription isodose to the target volume, multiplied by the ratio of the target volume receiving the prescription isodose to the prescription isodose volume; R50: ratio of volume receiving 50% of the prescription dose to the target volume; V12Gy: percent of brain volume receiving 12 Gy or more; F: forward planned; I: inversely planned.
sented by the R50 parameter, was shown to improve in ten of the eleven patients for inversely optimized plans (3.2 ± 0.7 and 3.9 ± 0.7 for inverse and forward plans, respectively). The V12Gy parameter, used here as a surrogate for normal brain complications, was also shown to improve in ten of the eleven patients for inversely optimized plans (4.2 ± 1.8% and 4.9 ± 2.1% for inverse and forward plans, respectively). Discussion and conclusion The CyberKnife and MultiPlan radiation delivery and treatment planning systems are shown in this work to be practical for multistage SRS of large cerebral AVMs. Highly conformal, composite treatment plans were successfully delivered to 11 patients with similar dose distributions as that of staged Gamma Knife treatments. The CI parameter is a measure of the compactness of the high dose distribution to an intended target as well as the location of the prescription isodose volume with respect to the target volume. A value close to 1.0 is expected to minimize radiation related complications. The average CI for the inverse planning technique in this study was shown to be comparable to that described for single fraction treatment of smaller AVMs using Gamma Knife Perfexion or Cyberknife systems (0.80) [13]. Dose-volume effects and parameters which correlate well with normal tissue complications for the staged treatment of large-volume AVMs have yet to be determined. The V12Gy index has been observed to correlate with the risk of developing SRS-related imaging changes and has been used as a surrogate for normal tissue damage [14]; thus the volume receiving 12 Gy is minimized in staged treatments, regardless of the number of stages used [15]. In this work, the composite V12Gy was less than 10% of the brain volume for all patients. This parameter continues to be used for plan evaluation and comparison as experience is gained in staged treatments of large AVMs. A new inverse optimization planning technique for staged treatments was shown to improve treatment plans in the majority of patients for all indices evaluated compared to forward optimized plans. The increased target conformality and normal tissue sparing allow for the possibility of increasing the prescription dose with the hope of improving nidus obliteration rates. The inverse planning technique developed for this technique is considerably more efficient than forward planning as the latter requires the treatment planner to envision a composite dose distribution while creating the dose distribution for individual sub-targets. This becomes increasingly difficult as the number of sub-targets is increased or for AVMs adjacent to organs at risk, such as the optic pathway.
With further patient follow-up, future work will investigate treatment outcome and toxicity for staged treatments of large AVMs using the CyberKnife delivery platform. Conflict of interest None. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.radonc. 2013.07.018. References [1] Lunsford LD, Niranjan A, Kondziolka D, et al. Arteriovenous malformation radiosurgery: a twenty year perspective. Clin Neurosurg 2008;55:108–19. [2] Flickinger JC, Kondziolka D, Maitz AH, Lunsford LD. An analysis of the doseresponse for arteriovenous malformation radiosurgery and other factors affecting obliteration. Radiother Oncol 2002;63:347–54. [3] Friedman WA, Bova FJ. Radiosurgery for arteriovenous malformations. Neurol Res 2011;33:803–19. [4] Kano H, Kondziolka D, Flickinger JC, et al. Stereotactic radiosurgery for arteriovenous malformations, Part 6: multistaged volumetric management of large arteriovenous malformations. J Neurosurg 2012;116:54–65. [5] Zabel A, Milker-Zabel S, Huber P, et al. Treatment outcome after linac-based radiosurgery in cerebral arteriovenous malformations: retrospective analysis of factors affecting obliteration. Radiother Oncol 2005;77:105–10. [6] Jayaraman MV, Marcellus ML, Do HM, et al. Hemorrhage rate in patients with Spetzler-Martin grades IV and V arteriovenous malformations – is treatment justified? Stroke 2007;38:325–9. [7] Barker FG, Butler WE, Lyons S, et al. Dose-volume prediction of radiationrelated complications after proton beam radiosurgery for cerebral arteriovenous malformations. J Neurosurg 2003;99:254–63. [8] Nagy G, Rowe JG, Radatz MWR, et al. A historical analysis of single-stage gamma knife radiosurgical treatment for large arteriovenous malformations: evolution and outcomes. Acta Neurochir 2012;154:383–94. [9] Back AG, Vollmer D, Zeck O, et al. Retrospective analysis of unstaged and staged Gamma Knife surgery with and without preceding embolization for the treatment of arteriovenous malformations. J Neurosurg 2008;109:57–64. [10] Adler JR, Chang SD, Murphy MJ, et al. The cyberknife: a frameless robotic system for radiosurgery. Stereotact Funct Neurosurg 1997;69:124–8. [11] Hristov D, Liu LN, Adler JR, et al. Technique for targeting arteriovenous malformations using frameless image-guided robotic radiosurgery. Int J Radiat Oncol 2011;79:1232–40. [12] Paddick I. A simple scoring ratio to index the conformity of radiosurgical treatment plans – technical note. J Neurosurg 2000;93:219–22. [13] Gevaert T, Levivier M, Lacornerie T, et al. Dosimetric comparison of different treatment modalities for stereotactic radiosurgery of arteriovenous malformations and acoustic neuromas. Radiother Oncol 2013;106:192–7. [14] Flickinger JC, Kondziolka D, Kalend AM, et al. Radiosurgery-related imaging changes in surrounding brain: multivariate analysis and model evaluation. Radiosurgery 1996;1:229–36. [15] Jones J, Jang S, Getch CC, et al. Advances in the radiosurgical treatment of large inoperable arteriovenous malformations. Neurosurg Focus 2007;23:E7.