International Journal of

Radiation Oncology

biology physics

www.redjournal.org

Imaging in Radiation Oncology

Velocity-based Adaptive Registration and Fusionfor Fractionated Stereotactic Radiosurgery Usingthe Small Animal Radiation Research PlatformPaul J. Black, PhD,* Deborah R. Smith, BS,* Kunal Chaudhary, PhD,*Eric P. Xanthopoulos, MD, JD,* Christine Chin, MD,*Catherine S. Spina, MD, PhD,* Mark E. Hwang, MD, PhD,*Mark Mayeda, MD,* Yi-Fang Wang, MS,* Eileen P. Connolly, MD, PhD,*Tony J.C. Wang, MD,*,y,z Cheng-Shie Wuu, PhD,* Tom K. Hei, PhD,*,x

Simon K. Cheng, MD, PhD,* and Cheng-Chia Wu, MD, PhD*

*Department of Radiation Oncology, Columbia University Medical Center, New York, New York;yDepartment of Neurological Surgery, Columbia University Medical Center, New York, New York;zHerbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, NewYork; xCenter for Radiological Research, Columbia University, New York, New York

Received Dec 30, 2017, and in revised form Apr 18, 2018. Accepted for publication Apr 23, 2018.

Summary

Fractionated magnetic reso-nance imaging (MRI)-guidedtreatment planning using theSmall Animal RadiationResearch Platform (SARRP)can be complex due to MRIregistration with cone beamcomputed tomography (CT)images. Using Velocity(Varian Medical Systems,Inc, Palo Alto, CA), the

Reprint requests to: Cheng-Chia Wu or Sim

of Radiation Oncology, Columbia University M

St, New York, NY 10032. Tel: (212) 305-707

org or sc3225@cumc.columbia.edu

Co-first authors: Paul J. Black and Debo

equally to this work.

Co-corresponding authors: Simon K. Che

contributed equally to this work.

Int J Radiation Oncol Biol Phys, Vol. -, No.

0360-3016/$ - see front matter � 2018 Elsevie

https://doi.org/10.1016/j.ijrobp.2018.04.067

Purpose: To implement Velocity-based image fusion and adaptive deformable regis-tration to enable treatment planning for preclinical murine models of fractionated ste-reotactic radiosurgery (fSRS) using the small animal radiation research platform(SARRP).Methods and Materials: C57BL6 mice underwent 3 unique cone beam computedtomography (CBCT) scans: 2 in the prone position and a third supine. A singleT1-weighted post-contrast magnetic resonance imaging (MRI) series of a murine met-astatic brain tumor model was selected for MRI-to-CBCT registration and gross tumorvolume (GTV) identification. Two arms were compared: Arm 1, where we performed3 individual MRI-to-CBCT fusions using rigid registration, contouring GTVs on each,and Arm 2, where the authors performed MRI-to-CBCT fusion and contoured GTVonthe first CBCT followed by Velocity-based adaptive registration. The first CBCT andassociated GTV were exported from MuriPlan (Xstrahl Life Sciences) into Velocity

on K. Cheng, Department

edical Center, 622 W 168

7; E-mail: chw9088@nyp.

rah R. Smith contributed

ng and Cheng-Chia Wu

Funding: This study was supported by the ASTRO 2016 Residents/

Fellows in Radiation Oncology Research Seed Grant, the Irving Institute

Imaging Pilot Award, Louis V. Gerstner, Jr. Scholar Award, and the NIH

Biomedical Research Support Shared Instrumentation Grant

1S10OD010631-01A1.

Conflict of interest: none.

Supplementary material for this article can be found at

www.redjournal.org.

-, pp. 1e7, 2018r Inc. All rights reserved.

Black et al. International Journal of Radiation Oncology � Biology � Physics2

authors have developed a

novel preclinical method formurine image fusion andadaptable registration tofacilitate fractionated radia-tion treatment planning. Tothe best of the authors’knowledge, this is the firststudy to apply clinical treat-ment planning software to-ward SARRP-basedtreatment planning toconduct translationalresearch using animals.

(Varian Medical Systems, Inc, Palo Alto, CA). In Arm 1, the second and third CBCTswere exported similarly along with associated GTVs (Arm 1), while in Arm 2, the first(prone) CBCT was fused separately to the second (prone) and third (supine) CBCTs,performing deformable registrations on initial CBCTs and applying resulting matricesto the contoured GTV. Resulting GTVs were compared between Arms 1 and 2.Results: Comparing GTV overlays using repeated MRI fusion and GTV delineation(Arm 1) versus those of Velocity-based CBCT and GTVadaptive fusion (Arm 2), meandeviations � standard deviation in the axial, sagittal, and coronal planes were0.46 � 0.16, 0.46 � 0.22, and 0.37 � 0.22 mm for prone-to-prone and0.52 � 0.27, 0.52 � 0.36, and 0.68 � 0.31 mm for prone-to-supine adaptive fusions,respectively.Conclusions: Velocity-based adaptive fusion of CBCTs and contoured volumes allowsfor efficient fSRS planning using a single MRI-to-CBCT fusion. This technique isimmediately implementable on current SARRP systems, facilitating advanced preclin-ical treatment paradigms using existing clinical treatment planning software. � 2018Elsevier Inc. All rights reserved.

Introduction

Radiation therapy plays an essential role in the manage-ment of brain metastases. With modern clinical advances inradiation therapy, stereotactic radiosurgery (SRS) hasemerged as standard-of-care treatment for select patientswith limited brain metastases (1, 2). SRS treatments aretraditionally delivered using a single fraction. However,both modern linear accelerator-based treatments and theLeksell Gamma Knife Icon (Elekta) allow for fractionatedSRS (fSRS) in appropriate patients (eg, for larger tumors orpostoperative cavities) who may benefit (3, 4, 5).

Given growing interest in expanding indications forSRS-based treatments, including potential combinationswith immunotherapy, preclinical models are needed toevaluate novel roles for single-fractionated and fractionatedSRS. The small animal radiation research platform(SARRP) (Xstrahl Life Sciences, Suwanee, GA) is a small-animal irradiator capable of advanced radiation treatmentplanning and delivery. Preclinical small animal models ofmagnetic resonance imaging (MRI)-guided radiation ther-apy using the SARRP have been successfully establishedfor intracranial, flank, and abdominal tumors using variousstrategies including usage of immobilization devices, fidu-cial marker placement, and MRI-only based treatmentplanning (6, 7, 8, 9, 10). While SARRP-based, MRI-guided, single-fraction SRS for brain tumors achievesprecise dose delivery (11), gross tumor volume (GTV)identification and treatment planning for fractionatedSARRP treatments currently requires repeat MRI fusionand registration to account for shifts in daily positioning,which is challenging and limits clinical applicability.Moreover, questions remain whether advanced clinicaltreatment planning technologies can be successfully inte-grated into existing workflows for small animal treatmentplanning to further advance sophistication and clinical

relevance of SARRP-based treatment delivery for preclin-ical models of radiation therapy.

Therefore, the purpose of this study was to implementadaptable registration using deformable image registrationwith Velocity (Varian Medical Systems, Inc, Palo Alto, CA)into existing workflows for SARRP-based small animaltreatment delivery to establish a preclinical murine modelfor SARRP-based fSRS. Here, the authors demonstrate thatVelocity-based cone beam computed tomography (CBCT)-to-CBCT image fusion and deformable registration cansuccessfully account for day-to-day treatment setup vari-ability and transform target volumes to allow for SARRP-based murine fSRS.

Methods and Materials

Experiments were conducted according to the authors’institutional animal care and use committee (IACUC).Seven-week-old male C57BL6 mice (Jackson Laboratory,Bar Harbor, ME) were anesthetized with xylazine andketamine. A representative post-contrast T1-weightedMRI scan of a mouse with a single brain metastasisfrom previous work was selected for MRI-to-CBCT rigidregistration (detailed information regarding generation ofthe orthotopic murine model provided in Wu et al.,IJROBP 2017) (11). Four mice each underwent 3 CBCTscans, 2 in the prone position (Scans A and B) and 1 su-pine (Scan C). To simulate fractionated treatment, micewere successively removed from the treatment stage andrepositioned before each CBCT scan. We investigatedboth prone-prone (A-B) and prone-supine (A-C) variationsto assess limitations of Velocity-based adaptive registra-tion. Manual MRI-to-CBCT rigid registration was per-formed in Muriplan (Xstrahl Life Sciences). A 5-pointpositioning system using bilateral orbits, auditory canals

Volume - � Number - � 2018 Velocity-based image fusion for SARRP 3

and the cribiform plate as anatomical landmarks, alongwith consideration of bony anatomy of the cranium, wasused to verify correct co-registration. CBCTs and associ-ated GTVs were exported from Muriplan into Velocity.Experiments were performed in duplicate by 2 indepen-dent users. Details regarding image and structure sethandling and transformations using custom-writtenMATLAB (MathWorks) scripts are provided in the Sup-plement. The authors compared correspondence of GTVsbetween 2 experimental arms (Fig. 1): Arm 1, where 3individual MRI-to-CBCT fusions were performed usingrigid registration with GTVs contoured on each CBCT,and Arm 2, where MRI-to-CBCT fusion and GTV con-touring were performed for only Scan A, using Velocity-based deformable registration with GTV adaptation forsuccessive CBCT scans (Scans B and C). Briefly, in Arm 2(Velocity-based adaptive registration), the first (prone)CBCT (Scan A) was separately fused to the second (ScanB) and third (Scan C) CBCTs. Deformable registrationswere performed on initial CBCTs (Scan A), and resultingtransformation matrices were applied to the corresponding

Scan B

Scan B Scan A + GTV(A)

Scan B + MRI fus

Adaptive Reg

Arm

1Ar

m 2

MRI-to-CBCT manual rigid registration (Mu

MRI-guided GTV delineation (Muriplan)

CBCT-TO-CBCT adaptive registration (Veloc

Adaption of GTV (Velocity)

1

1

2

3

3

4

Fig. 1. Overview of experimental design. In Arm 1 (blue box),and GTVs were contoured for each successive scan (depicted inGTV contouring were performed using Scan A. Velocity-based adto C were performed to generate Velocity-adapted GTVs for Sc

GTVs. Specifically, Scan B or C was registered as theprimary image and Scan A (along with GTV) as the sec-ondary image. Rigid registration was first performed withScan A to bring primary and secondary images into closeproximity, followed by deformable registration. Thetransformation matrix was then applied to the GTV fromScan A. The authors compared independently-contouredGTVs from Scans B and C in Arm 1 (manual MRI-CBCT rigid registration in Muriplan and GTV contour-ing on each scan) to Velocity-adapted GTVs applied tosuccessive CBCTs (Scans B and C) from Scan A in Arm 2by comparing GTV centroid shifts between methods in theaxial, sagittal and coronal planes. The related-samplesWilcoxon signed rank test was used to compare meanvector shifts of GTV centroids between prone and supinescans between Arm 1 and Arm 2 (eg, for target volumesrepeatedly contoured using manual CBCT-to-MRI rigidregistration vs Velocity-based adaptive fusion of GTVs). Pvalues less than .05 were considered statistically signifi-cant. Both intra-user and inter-user GTV variability werecharacterized using descriptive statistics.

Scan B

Scan B

Scan B

ion GTV(B) contour

GTV(B) Transformed

istration

riplan)

ity) with MRI and GTV fusion

2

4

individual MRI-to-CBCT rigid registrations were performedgreen). In Arm 2 (green Box), MRI-to-CBCT fusion andaptive registration of CBCTs from Scan A to B and Scan Aans B and C (red).

Fig. 2. Two independent CBCTs and GTVs before adaptive registration in Velocity.

Black et al. International Journal of Radiation Oncology � Biology � Physics4

Results

Velocity effectively performed all murine CBCT and GTVregistrations (Figs. 2 and 3). Figure 2 displays CBCT andGTV positions in Arm 2 before deformable registrationfrom Scan A (red) and Scan B (green). As shown inFigure 3, Velocity-based application of the associatedtransformation matrix to the GTV from Scan A resulted instrong co-localization with the corresponding GTV fromScan B. Even for successive CBCT scans with radical po-sition shifts (eg, prone-to-supine between Scans A and C),Velocity performed deformable registration with highcoincidence.

Deviations of the GTV centroids between Arm 1(manual MRI-to-CBCT and repeated GTV contouring)and Arm 2 (Velocity-based deformable registration) in all3 planes (axial, sagittal and coronal) were compared for alltransformed GTVs (Table 1). Scan A-to-B (prone-to-prone) average deviations between arms were 0.46 � 0.16,0.46 � 0.22, and 0.37 � 0.22 mm in the axial, sagittal andcoronal planes, respectively. Scan A-to-C (prone-to-su-pine) deviations were 0.52 � 0.27, 0.52 � 0.36, and0.68 � 0.31 mm, respectively. However, comparison ofArms 1 and 2 (manual CBCT-MRI rigid registration vsVelocity-based adaptive fusion) by comparing mean GTVcentroid vector shifts between prone and supine scans

Fig. 3. Two independent CBCTs and GTVs after adaptive registration in Velocity.

Volume - � Number - � 2018 Velocity-based image fusion for SARRP 5

across both methods revealed no statistically significantdifference between Arm 1 and Arm 2 (P Z .327). GTVcentroid variability (both inter-operator and intra-operator) are available in Tables E1 and E2, respectively(Tables E1 and E2; available online at www.redjournal.org). Dosimetric analysis examining overlaps in target

Table 1 Deviation of GTV centroid between Arm 1 and 2

Comparison Axial

Prone to prone (A to B) 0.46 � 0.16 0.4Prone to supine (A to C) 0.52 � 0.27 0.5

coverage between the 2 methods was also performed bydesigning an AP-PA plan to the Velocity-deformed GTVto evaluate relative coverage of the Muriplan MR-fusionderived GTV (Fig. E1; available online at www.redjournal.org). Overall, dosimetric analysis demon-strated high correspondence of GTV coverage between

Sagittal Coronal Total vector

6 � 0.22 0.37 � 0.22 0.78 � 0.092 � 0.36 0.68 � 0.31 1.05 � 0.15

Coronal Sagital Axial

CTCT

-MRI

MRI

-GTV

Adap

tive

GTV

Fig. 4. MuriPlan display of MRI-guided delineation of GTV (green) versus adaptive registration transformed GTV (red).

Black et al. International Journal of Radiation Oncology � Biology � Physics6

methods (Table E3; available online at www.redjournal.org). As demonstrated in Figure 4, final adaptiveregistration-transformed GTVs (red) were successfullyexported from Velocity and imported into MuriPlan fortreatment planning, which corresponded well withrepeated MRI-guided delineation of GTVs (green).

Discussion

This study demonstrates feasibility of Velocity-basedadaptation of SARRP treatment planning to allow for pre-clinical murine models of fractionated SRS. Throughdeformable registration and associated transformationmatrices, this technique successfully accounts for inter-fraction GTV position changes. Our results suggest that0.5 mm planning target volume expansions sufficientlyaccount for associated uncertainty. Given that multiple MRIacquisitions are both expensive and time-consuming, Ve-locity-based deformable registration of CBCT imagesprovide an appropriate alternative with excellent dosimetric

coverage for precise delivery of fractionated MRI-guidedtreatment delivery in small animal models. After initialmanual CBCT-to-MRI co-registration and GTV contouring,the current workflow for adaptive registration for succes-sive CBCT scans can be completed in approximately20 minutes. However, limitations include lack of fullautomatization of steps within our workflow (eg, CT in-tensity values of CBCTs from Muriplan must be convertedbefore Velocity importation,and transferring images be-tween software platforms still requires manual imageimport/export of images). Future work will include auto-mation of additional steps in our current workflow tostreamline this process (including creation of a direct linkbetween the Velocity database and 3DSlicer to efficientlypush images). As compared with prior studies of MRI-guided fractionated radiation delivery, this technique iseasily implementable using existing treatment planningsoftware and can also be used synergistically with immo-bilization devices and fiducials.

Radiation therapy holds exceptional promise towardenhancing the efficacy of immuno-oncology-based

Volume - � Number - � 2018 Velocity-based image fusion for SARRP 7

treatments (12). Both preclinical and clinical findingssuggest that radiation can induce abscopal effects throughimmune sensitization, and immune checkpoint inhibitionwith concurrent SRS may improve outcomes in the settingof brain metastases (13, 14, 15, 16). Preclinical models forboth accurate and clinically applicable delivery of radiationtherapy are essential for determining the optimal dose andfractionation schemes to elicit such effects (17).

This technique establishes a clinically-relevant preclin-ical model for murine MRI-guided fractionated SRS andalso offers many other exciting possibilities for advancingpreclinical radiation research, including positron emissiontomography/computed tomography (PET/CT) adaptivefusion for targeted therapy using spontaneous metastaticmouse models (18). Here, we demonstrate that Velocity-based adaptive registration can be easily incorporated intothe conventional Muriplan workflow. To our knowledge,this represents the first study to demonstrate feasibility ofapplying clinical image registration software towardSARRP-based murine treatment planning.

Reference

1. Citrin DE. Recent developments in radiotherapy. N Eng J Med 2017;

377:2200-2201.

2. Garrett MD, Wu CC, Yanagihara TK, et al. Radiation therapy for the

management of brain metastases. Am J Clin Oncol 2016;39:416-422.

3. Kim JW, Park HR, Lee JM, et al. Fractionated stereotactic gamma

knife radiosurgery for large brain metastases: A retrospective, single

center study. PloS One 2016;11:e0163304.

4. Stieler F, Wenz F, Abo-Madyan Y, et al. Adaptive fractionated ste-

reotactic gamma knife radiotherapy of meningioma using integrated

stereotactic cone-beam-CT and adaptive re-planning (a-gkFSRT).

Strahlenther Onkol 2016;192:815-819.

5. Wegner RE, Leeman JE, Kabolizadeh, et al. Fractionated stereotactic

radiosurgery for large brain metastases.Am JClin Oncol 2015;38:135-139.

6. Baumann BC, Benci JL, Santoiemme PP, et al. An integrated method

for reproducible and accurate image guided stereotactic cranial

irradiation of brain tumors using the small animal radiation research

platform. Transl Oncol 2012;5:230-237.

7. Bolcaen J, Descamps B, Deblaere K, et al. Mri-guided 3D conformal

arc micro-irradiation of a F98 glioblastoma rat model using the small

animal radiation research platform (SARRP). JNeurooncol 2014;120:

257-266.

8. Corroyer-Dulmont A, Falzone N, Kersamans V, et al. MRI-guided

radiotherapy of the SK-N-SH neuroblastoma xenograft model using a

small animal radiation research platform. Br J Radiol 2017;90:

20160427.

9. Gutierrez S, Descamps B, Vanhove C, et al. MRI-only based radio-

therapy treatment planning for the rat brain on a small animal radia-

tion research platform (SARRP). PloS One 2015;10:e0143821.

10. Kersemans V, Beech JS, Gilchrist S, et al. An efficient and robust

MRI-guided radiotherapy planning approach for targeting abdominal

organs and tumours in the mouse. PloS One 2017;12:e0176693.

11. Wu CC, Chaudhary KR, Na YH, et al. Quality assessment of stereo-

tactic radiosurgery of a melanoma brain metastases model using a

mouse-like phantom and the small animal radiation research platform.

Int J Radiat Oncol Biol Ph 2017;99:191-201.

12. Antonia SJ, Villegas A, Daniel D, et al. Durvalumab after chemo-

radiotherapy in stage III non-small cell lung cancer. N Engl J Med

2017;377:1919-1929.

13. Demaria S, Coleman CN, Formenti SC, et al. Radiotherapy: Changing

the game in immunotherapy. Trends Cancer 2016;2:286-294.

14. Goldberg SB, Gettinger SN, Mahajan A, et al. Pembrolizumab for

patients with melanoma or non-small cell lung cancer and untreated

brain metastases: Early analysis of a non-randomised, open-label,

phase 2 trial. Lancet Oncol 2016;17:976-983.

15. Tallet AV, Dhermain F, Le Rhun E, et al. Combined irradiation and

targeted therapy or immune checkpoint blockade in brain metastases:

Toxicities and efficacy. Ann Oncol 2017;28:2962-2976.

16. Weichselbaum RR, Liang H, Deng L, et al. Radiotherapy and

immunotherapy: A beneficial liaison? Nat Rev Clinl Oncol 2017;14:

365-379.

17. Vanpouille-Box C, Alard A, Aryankalayil MJ, et al. DNA exonuclease

trex1 regulates radiotherapy-induced tumour immunogenicity. Nat

Commun 2017;8:15618.

18. Cho JH, Robinson JP, Arave RA, et al. AKT1 activation promotes

development of melanoma metastases. Cell Rep 2015;13:898-905.