effect of anatomic motion on proton therapy dose distributions in prostate cancer treatment

P

TgtXrtctamivpCcn

U1F

Int. J. Radiation Oncology Biol. Phys., Vol. 67, No. 2, pp. 620–629, 2007Copyright © 2007 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2006.10.008

HYSICS CONTRIBUTION

EFFECT OF ANATOMIC MOTION ON PROTON THERAPY DOSEDISTRIBUTIONS IN PROSTATE CANCER TREATMENT

XIAODONG ZHANG, PH.D.,* LEI DONG, PH.D.,* ANDREW K. LEE, M.D.,† JAMES D. COX, M.D.,†

DEBORAH A. KUBAN, M.D.,† RON X. ZHU, PH.D.,* XIAOCHUN WANG, PH.D.,* YUPENG LI, M.S.,*WAYNE D. NEWHAUSER, PH.D.,* MICHAEL GILLIN, PH.D.,* AND RADHE MOHAN, PH.D.*

Departments of *Radiation Physics and †Radiation Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, TX

Purpose: To determine the dosimetric impact of interfraction anatomic movements in prostate cancer patientsreceiving proton therapy.Methods and Materials: For each of the 10 patients studied, 8 computed tomography (CT) scans were selectedfrom sets of daily setup CT images that were acquired from a cohort of prostate cancer patients. The images wereacquired in the treatment room using the CT-on-rails system. First, standard proton therapy and intensity-modulated radiation therapy (IMRT) plans were designed for each patient using standard modality-specificmethods. The images, the proton plan, and the IMRT plan were then aligned to the eight CT images based onskin marks. The doses were recalculated on these eight CT images using beam from the standard plans. Second,the plans were redesigned and evaluated assuming a smaller clinical target volume to planning target volumemargin (3 mm). The images and the corresponding plans were then realigned based on the center of volume ofthe prostate. Dose distributions were evaluated using isodose displays, dose–volume histograms, and targetcoverage.Results: For the skin-marker alignment method, 4 of the 10 IMRT plans were deficient, whereas 3 of 10 protonplans were compromised. For the alignment method based on the center of volume of the prostate, only theproton plan for 1 patient was deficient, whereas 3 of the 10 IMRT plans were suboptimal.Conclusion: A comparison of passively scattered proton therapy and highly conformal IMRT plans for prostatecancer revealed that the dosimetric impact of interfractional anatomic motions was similar for both modalities.© 2007 Elsevier Inc.

Proton therapy, Image-guided radiotherapy, Intensity-modulated radiation therapy, Organ motion.

scttssptmptTfasi

I

A

INTRODUCTION

wo recent technologic advances have the potential toreatly improve the field of radiation therapy: proton beamherapy and image guidance. Clinical proton beams, unlike-ray beams, have a low entrance dose, followed by a

egion of uniform high dose (the spread-out Bragg peak) athe tumor/target, then a steep falloff to zero dose (1). Theseharacteristics make possible a substantial dose reduction tohe normal tissues while maximizing the dose to the tumornd give proton therapy an inherent advantage over confor-al photon therapy (2–4). Simultaneous with increasing

nterest in proton therapy, there has been a significant ad-ancement in imaging techniques (5–9). For example, re-eat computed tomography (CT) imaging using in-roomT-on-rails or cone-beam CT have become available fororrecting interfractional setup errors or for adaptive replan-ing (8, 10–12).

Reprint requests to: Xiaodong Zhang, Ph.D., Radiation Physics,nit 94, The University of Texas M. D. Anderson Cancer Center,515 Holcombe Blvd., Houston, TX 77030. Tel: (713) 563-2533;ax: (713) 794-2545; E-mail: [email protected]
Supported in part by grant CA74043 from the National Cancer
620

Repeat imaging is important because the locations,hapes, and sizes of diseased tissue and normal anatomy canhange significantly because of daily positioning uncertain-ies and anatomic changes during the course of radiationreatments as a result of nonrigidity of the body, tumorhrinkage, weight loss, and variations in anatomic contents,uch as rectal gas and bladder filling in prostate canceratients (9–11, 13, 14). Because of these changes, thehree-dimensional (3D) CT images used for radiation treat-ent planning do not necessarily correspond to the actual

osition of the anatomy at the time of delivery of eachreatment fraction or even to the mean treatment position.he traditional assumption—that the anatomy discerned

rom 3D CT images acquired for planning purposes ispplicable for every fraction—does not take into accountuch interfractional changes and may ultimately limit the abil-ty to fully exploit the potential of external beam radiotherapy.

nstitute.Conflict of interest: none.Received July 6, 2006, and in revised form Sept 28, 2006.

ccepted for publication Oct 2, 2006.

hddsbsptdfpktsoir

immccoI

P

odIsttmptrieisat1tvCaitd

T

i

ptssc2smwuisTsmmdaa3ap(

Tpimt

Ts2tdaeenttdmu

pcc(pimd

621Effect of anatomic motion on proton therapy dose distributions ● X. ZHANG et al.

For these reasons, there is concern that highly conformal,igh-dose intensity-modulated radiation therapy (IMRT)ose distributions designed on the basis of a single CTataset acquired for planning purposes may lead to unfore-een complications or to marginal misses of target volumesecause of the interfractional movement of the normaltructures and the target volumes. It has been argued thatroton therapy is more susceptible to tissue density uncer-ainties than photon therapy (15). Numerous studies haveemonstrated changes in the 3D dose photon distributionrom interfractional variations in the shapes, sizes, andositions of anatomic structures (9–11, 13, 14). To ournowledge, however, there have been no similar investiga-ions for proton therapy. With access to the CT-on-railsystem and proton and IMRT treatment planning systems inur institution, we are able to investigate the dosimetricmpact of interfractional movement of anatomy in patientseceiving proton therapy and IMRT.

The goal of this study was to assess the dosimetric effectsn prostate cancer patients caused by interfractional move-ent of anatomy in the path of the proton beam. To deter-ine how the 3D dose distributions changed during the

ourse of proton therapy, we undertook a retrospectiveomparative treatment planning study using repeat CT databtained from 10 prostate cancer patients who had receivedMRT treatments at our institution.

METHODS AND MATERIALS

atient dataTen prostate cancer patients previously irradiated by photons at

ur institution were selected for this study. All patients had aiagnosis of early-stage prostate carcinoma (T2a) and had receivedMRT in a linear accelerator suite equipped with the CT-on-railsystem (10, 11). Images for each patient were acquired two orhree times per week using the CT-on-rails system just before thereatment. The immobilized patient was positioned on the treat-ent couch, where the couch was in approximately the treatment

osition. The couch was then rotated 180° to allow the CT-on-railso move into position and acquires an image. The couch was thenotated back to the treatment delivery position. In addition to thenitial CT scan acquired for treatment planning, 24 CT scans forach patient were acquired over the course of radiotherapy thatncluded 42 treatment fractions in approximately 8 weeks. For thistudy, for practical reasons, we selected a subset of 8 CT scans,pproximately 1 scan per week, for each patient. Figure 1 showshe corresponding CT sections for the eight fractions selected forof the patients. The prescribed dose for each patient was 75.6 Gy

o 98% of the planning target volume (PTV). The clinical targetolume (CTV) included the prostate and the seminal vesicles. TheTV-to-PTV margin for photon IMRT planning was 8 mm, exceptt the rectum-prostate interface, where it was 5 mm. The repeat CTmages were used to study interfractional anatomic changes andheir dosimetric consequences, not for the modification of actualaily treatment. Contours on all CT scans were drawn manually.

reatment planning with initial CT imagesA commercial treatment planning system (Eclipse; Varian Med-

cal Systems, Inc., Palo Alto, CA) was used for both IMRT and f

roton therapy plans. The proton beam commissioning data for thereatment planning system were obtained from the Monte Carloimulations, which showed very good agreement with the mea-ured data. For IMRT planning, each patient plan included eightoplanar, 6-MV beams placed at gantry angles of 25, 60, 100, 150,20, 260, 300, and 335° (International Electro Technical Commis-ion scale) were used. The Eclipse inverse treatment planningodule generated optimized photon intensity distributions, whichere then used to generate leaf sequences, which were in turnsed to compute the dose distribution for the planning CTmages. The proton therapy plans were designed using thetandard Loma Linda approach with two lateral beams (16, 17).he key parameters for proton therapy plans are “bordermoothing,” “smearing,” aperture margins, and distal and proximalargins. Most of the planning parameters are selected using theethods described by Moyers et al. (18, 19). The compensator was

esigned for the CTV using a custom distal margin that included3.5% of depth to account for uncertainty for CT number accuracynd conversion to proton relative linear stopping power and a-mm range uncertainty to take into account uncertainties in theccelerator energy, variable scattering system thickness, and com-ensator density. The distal margin (DM) and proximal marginPM) are given in Eq. 1.

DM � 0.035 � CTV distal depth � 3 mm

PM � 0.035 � CTV distal depth � 3 mm(1)

he aperture margins (AM) for all proton beams were drawn toroject outside of the CTV by a distance corresponding to thenternal target motion margin (IM) plus the setup uncertainty

argin (SM) plus the 90%–50% penumbral width as determined athe widest part of the target, or

AM � IM � SM � (90% � 50% penumbra) (2)

he planning system designed the range compensators using aimple ray tracing from the virtual proton source position (about70 cm from isocenter) through the CTV. If the compensatorhickness outside of the region included in the ray tracing proce-ure were set to the maximum thickness, protons would scrapelong the walls of the ray-traced region of the compensator wher-ver the compensator is not shielded by the block. To avoid thisffect, a “border-smoothing” margin was specified in the plan-ing system to set the compensator thickness t not shielded byhe block to the average thickness of the compensator traced byhe ray and located within a circle centered at t with a radiusefined by the border-smoothing margin. The border-smoothingargin is not a critical parameter, and the default value (1 cm) was

sed.In proton therapy, uncertainty in aligning the compensator to the

atient and possible motion of the patient during treatment canreate cold spots in the target. To guarantee target coverage, theompensator was “smeared” using the algorithm from Urie et al.20) and implemented in the treatment planning system. It com-ares each pixel of the ray tracing compensator with the pixelsnside the smearing margin and sets the pixel value to theinimum of the neighboring pixels. In our proton therapy plan

esign, the smearing parameter was selected on the basis of the
ormula by Moyers et al. (18), or

wpdspp1ociitose

tocct

A

Crvltatms

n show

622 I. J. Radiation Oncology ● Biology ● Physics Volume 67, Number 2, 2007

Smearing ��(IM � SM)2 � [0.03 � (distal CTV depth

� compensator thickness)]2, (3)

here IM is the internal margin and SM is the setup margin. Forrostate plans, we set the IM � 0. The SM was set to valueserived from the CTV-to-PTV expansion. For all the fields in thistudy, the distal CTV depth was about 24.4 � 0.8 cm and theroximal CTV depth was about 15.2 � 1.5 cm. A typical com-ensator thickness for prostate treatments with our system is 5.8 �.1 cm of Lucite. Combining these values and evaluating Eq. 3, webtained a smearing parameter value (radius) of approximately 1.2m for a typical prostate case. It is noteworthy that Eq. 3 actuallymplies that smearing plays two roles. First, it compensates forntrafractional and interfractional variations in tissue densities inhe path of the beam. Second, it mitigates against the consequencesf the approximations in the design of compensators. In the currenttate of the art, the water-equivalent compensator thickness along

Fig. 1. Computed tomography images acquired on eight tcoronal views). The images were aligned as if the patienthe prostate, the red contour shows the rectum, and cya

ach ray is set to the water-equivalent thickness through the tissues c

o the distal edge of the target plus the margin. The lateral transportf radiation is ignored. However, when the dose distribution isalculated using such a compensator, defects in the form of hot andold spots may appear near the end of the range. Smearing serveso minimize such defects.

nalysis of repeated CT imagesThe proton therapy and IMRT plans designed using the planning

T images were aligned with each of the selected subset of eightepeat CT images based on the skin marks and on the center ofolume of the prostate. The dose distributions were then recalcu-ated for the repeat CT images using the same beam portals (i.e.,he same beam range, spread-out Bragg peak (SOBP) width,perture, range compensators, and normalization for the protonherapy plans and the same energy, multileaf collimator leaf–otion patterns for the IMRT plans). For plans computed using the

kin-marker alignment, a CTV-to-PTV margin of 5 or 8 mm,

nt fractions for a prostate cancer patient (transverse andsetup using skin landmarks. The red color wash showss the bladder.

reatmet were

urrently used at M. D. Anderson for IMRT, was used for plan

eaua

eitpavIiIcrahf

C

fFpt

PpCct

Fcpcb


valuation. For the plans using the prostate center of volumelignment method, a uniform 3-mm CTV-to-PTV margin wassed. We term the plans recalculated using the repeat CT imagess repeat CT plans.

Because contours on repeat CT images had already been delin-ated, for this study, we used the average volumes receiving thendicated doses or higher for the rectum, bladder, and CTV overhe eight fractions to approximate the dose and volume data for thelan delivered during the course of the treatment. We termed theverage volume receiving the indicated dose as the “recalculatedolume receiving the delivered dose” in the remaining discussion.t should be noted that using the average volumes receiving spec-fied doses over multiple fractions may not be a valid concept.ndividual voxels change position and, in some situations, volumesontained within voxels may also change. Ideally, deformableegistration should be used to track individual voxels and compute

biologically equivalent dose for each voxel. The data shownere, however, are sufficient to give a qualitative sense of theraction-to-fraction variability of dose–volume combinations.

RESULTS

omparison of IMRT and proton therapy plansTypical dose distributions in transverse and sagittal planes

or the original IMRT and proton therapy plans are shown inig. 2. All dose distributions were normalized to 98% of therescribed dose to the PTV (75.6 Gy). Figure 3 compareshe corresponding dose–volume histograms for the CTV,

Fig. 2. The dose distributions in a transverse plane
intensity-modulated radiation therapy plan (left panel) and a p
TV, bladder, femoral heads, and the rectum. The IMRTlan had slightly lower dose homogeneity in the PTV andTV as a consequence of conformal avoidance of nearbyritical structures (i.e., the rectum and bladder). The protonherapy plan was better at sparing the rectum at doses of less

anel) and in a sagittal plane (bottom panel) for an

ig. 3. Dose–volume histograms for the planning target volume,linical target volume, rectum, femoral heads, and bladder for arostate cancer case. The solid and dashed lines show the resultsalculated using intensity-modulated radiation therapy and protoneams, respectively.

(top p
roton therapy plan (right panel).

tsPfiftbcdm1wb1Idittta

cIapVa5Tcpd

bvsdCgpvrtmt

t5sftSa4wOsp

vbdpuuatr

R

B

C

CC

ao3

R

B

C

aTt


han 50 Gy. However, above 50 Gy, IMRT was better atparing the rectum. The body mean nontarget (excludingTV) integral dose was 6 Gy for the IMRT plan and 3.6 Gyor the proton therapy plan, indicating that the nontargetntegral dose was 1.7 times higher for the IMRT plan thanor the proton therapy plan. The proton therapy plan sparedhe bladder better than IMRT plans for doses below 45 Gy,ut at higher doses the two plans were very similar. We alsoalculated the conformality index (PTV volume/prescribedose volume) for the patient shown in Fig. 2. The confor-ality index for the IMRT plan is 1.16 and proton plan is

.33. The main reason for large lateral treatment volumeas that we selected the distal margin (left-right direction)ased on the 3.5% CT number uncertainties, which is about.2 cm larger than the 0.3-cm margin chosen for the photonMRT plan in left-right direction. Figure 3 illustrates theifferences between an IMRT and a proton plan. This cases representative of the 10 cases studied in this work. Onhe basis of dose considerations, these results suggest thathe proton plan provides a significantly lower integral doseo healthy tissues and comparable target coverage andvoidance of critical structures.

An analysis of the irradiated tissue volumes generallyonfirmed the results of dose analysis described previously.n Tables 1 and 2, we present average volumes of the rectumnd bladder receiving at least the indicated doses with theroton therapy plans (column III) and IMRT plans (column) using a uniform 3-mm CTV-to-PTV margin (Table 1)

nd a larger margin (8-mm CTV-to-PTV margin except-mm margin at rectum and prostate interface) (Table 2).he volumes were averaged over 10 patients. Similar to thease shown in Figs. 2 and 3, at doses less than 50 Gy, theroton therapy plan was superior in sparing the rectum. Foroses higher than 50 Gy, IMRT plans spared the rectum

Table 1. Prostate dose–volume data averaged over 8 fractionsand 10 patients

ROIDose(Gy)

Pre-RT(proton)

CT-guidedregistration

(proton)Pre-RT(IMRT)

CT-guidedregistration

(IMRT)

ectum 30 47.8 52.4 61.7 64.540 39.7 45.3 46.8 55.450 32.1 38.3 33.4 40.860 23.7 30.8 21.5 28.870 13.8 21.6 9.5 17.4

ladder 30 12.9 23.7 21.8 42.540 10.9 20.1 15.8 31.650 9.0 16.6 11.6 23.260 7.1 13.1 8.0 16.070 4.8 8.9 5.2 10.3

TV 75.6 100.0 98.2 100.0 95.6

Abbreviations: ROI � region of interest; RT � radiotherapy;T � computed tomography; IMRT � intensity-modulated RT;TV � clinical target volume.The data are for the original plan and for the original plan

pplied to subsequent fractions with setup based on prostate centerf volume. The original plans were designed using a uniform

t-mm planning target volume to PTV expansion margin.

etter. We also calculated the differences in normalizedolume for the rectum between the plans using larger andmaller margins at various doses. The data show a potentialosimetric benefit with smaller margins when using theT-guided techniques for patient setup. Reducing the mar-in size resulted in more rectal sparing improvements in theroton therapy plans than those on the IMRT plans. Theolumes receiving doses of 30, 40, 50, 60, and 70 Gy wereeduced by 8.7%, 9.3%, 9.2%, 8.9%, and 7.7%, respec-ively, for the proton therapy plans when the CTV-to-PTVargin was reduced to 3 mm; the corresponding values for

he IMRT plan were 5.4%, 4.7%, 5.1%, 5.7%, and 6.0%.For the bladder, at doses lower than 50 Gy, the proton

herapy plan produced superior results. At doses higher than0 Gy, IMRT plans and proton therapy plans had similarparing of the bladder. Reducing the CTV-to-PTV marginrom 8 mm to 3 mm resulted in a larger reduction in dose tohe bladder in proton therapy plan than in the IMRT plan.pecifically, the volumes receiving doses of 30, 40, 50, 60,nd 70 Gy were reduced by 5.1%, 4.9%, 4.7%, 4.5%, and.0% for the proton therapy plans; the corresponding valuesere 5.1%, 4.6%, 4.1%, 3.7%, and 3.3% for the IMRT plan.n average, reducing the margin had less effect on bladder

paring than on rectal sparing for both proton and IMRTlans.Table 1 also presented the recalculated volume receiving

arious dose or higher for the rectum, bladder, and CTV foroth IMRT and proton repeat CT plans. The table comparesose and volume data (averaged over 10 patients) from theretreatment proton and IMRT plans. The recalculated vol-me data were averaged over the eight fractions using aniform 3-mm CTV-to-PTV margin and the CT-guidedlignment method. Similar to the original plan, at doses lesshan 50 Gy, the proton therapy plan produced superioresults. For doses higher than 50 Gy, the IMRT plans spared

Table 2. Prostate dose–volume data averaged over 8 fractionsfor the 10 patients

ROIDose(Gy)

Pre-RT(proton)

Skin markregistration

(proton)Pre-RT(IMRT)

Skin markregistration

(IMRT)

ectum 30 56.5 54.1 67.1 71.040 49.0 47.5 51.5 55.250 41.2 40.8 38.5 40.960 32.6 33.4 27.2 29.470 21.6 24.2 15.5 19.4

ladder 30 18.1 31.8 26.9 48.740 15.9 28.3 20.4 38.550 13.8 25.0 15.7 30.660 11.6 21.3 11.7 23.270 8.8 16.6 8.5 17.4

TV 75.6 100.0 96.3 96.2 94.5

Abbreviations as in Table 1.The data are for the original plan and for the original plan

pplied to subsequent fractions with setup based on skin marker.he original plans were designed using a larger CTV-to-planning

arget volume margin currently used at M. D. Anderson for IMRT.

he rectum better. From the results in Table 1, we calculated

tpdbtt7bat

tctva6fvIvpp

mudibav

tepia

Tr

oeatstpafdIldn(mpdecp9


he differences between the original plans and repeat-CTlans in the normalized volume of the rectum that receivedoses of 30, 40, 50, 60, and 70 Gy. The volumes increasedy 4.6%, 5.6%, 6.3%, 7.1%, and 7.8%, respectively, fromhe original plans to the repeat-CT plans for proton therapy;he respective values were 2.9%, 8.6%, 7.4%, 7.3%, and.9% for the IMRT plans. At most doses, the differencesetween the normalized volumes for the repeat CT plansnd those for the original plans were smaller for protonherapy than for IMRT.

Repeat-CT proton therapy plans spared the bladder betterhan the repeat-CT IMRT plans for all dose levels. We alsoalculated the differences between normalized volume forhe original plans and the repeat CT plans in the normalizedolume of the bladder that received doses of 30, 40, 50, 60,nd 70 Gy. The volumes increased by 10.8%, 9.1%, 7.5%,.0%, and 4.1%, respectively, for the proton therapy plansrom the original plans to repeat CT plans; the respectivealues were 20.7%, 15.7%, 11.7%, 8.0%, and 5.1% for theMRT plans. At all doses, the differences between theolumes of the repeat CT plans and volumes for the originallans were smaller for the proton plan than for the IMRTlan.For the plans using a larger margin (8 mm CTV-to-PTVargin except 5 mm at rectum and prostate interface), we

sed the skin-mark alignment method to calculate the actualose–volume data. Table 2 presents volumes receiving var-ous doses or higher for the rectum, bladder, and CTV foroth IMRT and proton repeat-CT plans. The values wereveraged over the 10 patient cohorts. A comparison of thesealues to the corresponding dose volume data for the pre-

-1-1

-0.5

0

0.5

1

1.5

0 1 2 3 4 5 6 7 8 9 10 1

PatPatient #

dx (c

m)

dx (c

m)

-1-1

-0-0.5

0

0.5

1

1.5

0 1 2 3 42 3 4 5 6 7 86 7 8 9 10 11

Patient #

dy (c

m)

dy (c

m)

(a)(a)

(b)(b)

Fig. 4. The mean differences between the center of vofractions for the 10 patients in (a) the right-left directiodirection. The error bars indicate the standard deviatio
directions.
reatment proton and IMRT plans revealed that the differ-nces between the original CT-based plans and the re-eat-CT plans were similar to the corresponding differencesn dose volume data using the CT-guided alignment methodnd a smaller CTV-to-PTV margin.

arget dose distribution for proton and IMRTepeat-CT plans

Figure 4 plots the average differences between the centerf volume of the prostate and the skin marker position overight fractions for the 10 patients in the right-left (Fig. 4a),nteroposterior (Fig. 4b), and superoinferior (Fig. 4c) direc-ions. The error bars indicate the standard deviation. Figure 4dhows the magnitude of the differences in the three direc-ions. The difference between the center of volume of therostate and the skin marker position was largest in thenteroposterior direction. This difference exceeded 0.5 cmor 3 patients (Patients 2, 4, and 9). In the superoinferiorirection, the difference exceeded 0.5 cm only for Patient 9.n the right-left direction, the difference for all patients wasess than 0.5 cm. Figure 4d shows that the magnitude of theifference was the largest for Patient 9. Figure 5a shows theormalized CTV volume receiving the prescribed dose75.6 Gy) averaged over the eight fractions using the skin-arker alignment method for both the IMRT and proton

lans. If 97% of the CTV or more receives the prescribedose, the plan is considered acceptable, otherwise it is consid-red to be unacceptable because of compromised targetoverage. Using the skin-mark alignment technique, theroton therapy plans for 3 of 10 patients (Patients 2, 4, and) were unacceptable. These patients exhibited larger dif-

-1-1

-0-0.5

0

0.5

1

1.5

0 10 1 2 3 4 5 65 6 7 8 9 10 11

PatPatient #

0

0.2

0.4

0.6

0.8

1

0 1 2 3 4 5 6 7 8 9 10 11

Patient #

(c)(c)

(d)(d)

f the prostate and skin marker position over the eightthe anteroposterior direction, and (c) the superoinferior

shows the magnitude of the differences in the three

1

11

dz (c

m)

dz (c

m)

sqrt

(dx^

2+dy

^2+d

z^2)

sqrt

(dx^

2+dy

^2+d

z^2)

lume on, (b)n; (d)

fsooebt(csp9

ttoup(adpa

pwm

satmpddeqpotteweceaiowitiiitd

py and


erences between the center of volume of the prostate andkin marker position in the anteroposterior direction thanther patients did. When the difference between the centerf volume of the prostate and the skin marker positionxceeded 5 mm, which is the smallest CTV-to-PTV margin,oth the proton and IMRT plans would have compromisedarget dose coverage. For the skin-marker alignment method, 4Patients 2, 3, 4, and 9) of the 10 IMRT plans were unac-eptable. The proton therapy plans are more tolerant of dailyetup variations. The average CTV volume receiving therescribed dose was 96.3% for the proton therapy plans and4.1% for the IMRT plans.Figure 5b shows the normalized CTV volume receiving

he prescribed dose (75.6 Gy) averaged over the eight frac-ions with repeat CT scans aligned using the center of volumef the prostate. The original plans were designed using aniform 3-mm CTV-to-PTV margin. For the proton therapylans, only the plan for Patient 9 is unacceptable, whereas 3Patients 1, 3, and 9) of the 10 IMRT plans are unaccept-ble. Here also the proton therapy plans are more tolerant ofaily setup variations. The average volume receiving therescribed dose for CTV was 98.2% for the proton plansnd 95.6% for the IMRT plans.

DISCUSSION

There are numerous differences between proton- andhoton-beam therapy that have significant consequencesith respect to planning and delivery of treatments. Com-

8 58 78 99 19 39 59 79 9

1 0 1

P a t ie n t1

P a t ie n t2

P a t ie n t3

P a t ie n t4

P

V Pr

escr

iptio

n

7 58 08 59 09 5

1 0 01 0 5

Patien

t 1

Patien

t 2

Patien

t 3

Patien

t 4

Patien

t 5

V Pr

escr

iptio

n

Fig. 5. Recalculated normalized clinical target volume refor the 10 patients using (a) the skin-marker alignmentprostate for both the intensity-modulated radiation thera

on to both modalities, however, is an incomplete under- p

tanding of the uncertainties in the dose distributions thatre actually delivered to prostate therapy patients, especiallyhose uncertainties introduced by interfraction anatomicotions. Protons have a finite range and a monoenergetic

roton beam is expected to virtually stop at a well-definedepth when incident normally on a flat homogeneous me-ium. Uncertainties related to CT numbers, stopping pow-rs, motion, and positioning affect protons and photonsuite differently. As a consequence, the depth at which therotons really stop is uncertain. Furthermore, the presencef inhomogeneities and compensators may degrade the pro-on range significantly. At the same time, component ofranslational motion for the body as a whole has virtually noffect on proton dose distributions; however, variation inater-equivalent path length to the proximal and distal

dges of the CTV must be accounted for. In addition, asompared with photons, protons scatter significantly differ-ntly from inhomogeneities and other objects in their pathnd may create hot and cold spots in regions distal tonhomogeneities. Moreover, commonly used algorithms cannly approximately account for proton scattering. Thushat is seen on a treatment plan is an approximation of what

s actually delivered. Decades of experience with protonherapy have resulted in empirical strategies to minimize thempact of these uncertainties. One such strategy is the wayn which margins are chosen. For photons in which a PTVs defined to account for positioning and motion uncertain-ies and margins of all beams are set to produce target doseistributions that adequately cover the PTV. In contrast, for

t P a t ie n t6

P a t ie n t7

P a t ie n t8

P a t ie n t9

P a t ie n t1 0

to n IM R T

nt 6

Patien

t 7

Patien

t 8

Patien

t 9

Patien

t 10

to n IM R T (a)

(b)

g the prescribed dose (75.6 Gy) over the eight fractionsand (b) the method using the center of volume of theproton repeat-computed tomography plans.

a t ie n5

P r o

Patie

P r o

ceivinmethod

rotons, margins must be defined for each beam. The prox-

ipfsctmtspec

sobbapulpaafdr

li

rmplptdnpd9pdetdpriw

stW


mal and distal margins are set so as to cover the CTV in theresence of uncertainties in the range of protons caused byactors mentioned previously (Eq. 1). Surprisingly, our re-ults suggest that that when the difference between theenter of volume of the prostate and the skin marker posi-ion exceeded 0.5 cm, which is the smallest CTV-to-PTVargin, both the proton and IMRT plans would have caused

he target to be compromised. This implies that major rea-on for compromising the target coverage for the protonlan in this work is setup errors, rather than proton rangerrors, would be the predominant cause of inadequate targetoverage in prostate patients planned.

Moyers et al. (18) argued that use of the PTV concepthould be abandoned for the charged particle beams basedn the treatment plan design for the lung patients. However,ased on this work, we would argue that PTV concept coulde used in the special case of a lateral opposed-pair fieldrrangement for treatment of the prostate. To illustrate thisoint, we designed a study to show the effect of the setupncertainty and range uncertainty on the prescription doseine. Figure 6 shows the original prescription line, and therescription lines when the patient was moved 5 mm each innterior, posterior, superior, inferior, right, and left directionnd stopping power for the tissue increased/decreased 3.5%or a parallel two-beam proton plan from right and leftirections. In the directions perpendicular to the beam di-ection, if the patient moves 5 mm, the prescription isodose

Fig. 6. The prescription dose line, and the prescription d(green), posterior (cyan), superior (pink), inferior (orangfor the tissue increased (black)/decreased (white) 3.5% foThe color washed blue shows the clinical target volum
expanded uniformly 5 mm.
ine will shift 5 mm accordingly. This is very similar to thesodose line change for a photon plan.

However, for the proton plan, we must also consider theange uncertainty along the beam direction. If the patientoves 5 mm along the beam directions, the corresponding

rescription dose lines were identical to the original doseine. This behavior is unique to the parallel opposed-pairroton beam arrangement. From Fig. 6, indications are thatarget is well covered even if the range is increased orecreased 3.5%. Although the PTV for the proton plan isot exactly the same as that for the photon plan, if therescription requires 98% of PTV receiving prescriptionose for the photon plan, it is also convenient to require8% photon PTV receiving prescription dose for the protonlan to ensure the target coverage for the direction perpen-icular to the beam directions. In a recent study, Thomast al. (21) showed that the PTV margin can be reduced inhe axial direction, but no reduction can be seen in otherirections for prostate proton plans compared with photonlans. Although Thomas et al.’s study did not consider theange uncertainties, the estimation of the setup uncertaintiesn the direction perpendicular to the beam directions agreeith our results.In the design of the proton plan, it is essential to use

mearing to compensate for intrafractional and interfrac-ional variations in tissue densities in the path of the beam.

ith the use of these margins and with the use of smearing,

es when the patient was moved 5 mm each in anteriort (blue), and left (brown) direction and stopping powerllel two-beam proton plan from right and left directions.color washed green shows the clinical target volume

ose line), righr a parae and

wtc

fctsstwpstppdrHwp

p

cs


e found that the proton therapy plans are no more sensitiveo interfractional variations in anatomy than the highlyonformal IMRT plans.

Fig. 7. Dose distributions in transverse planes for an intbeam with four beams, (b) a proton plan with two phistograms, (d) for the planning target volume (PTV), rdashed lines show the results calculated using four ob

Table 3. The CTV, prostate, and seminal vesicle volumesreceiving the prescribed dose (75.6 gy) for the 8 daily CT plans

for both proton and IMRT plans for Patient 9

DailyCT #

IMRT Proton

CTV SV Prostate CTV SV Prostate

1 86.5 53.7 97.2 92.7 76.5 97.82 88.1 56.5 98.8 94.0 78.6 99.13 87.6 57.7 98.0 95.1 84.3 98.84 88.0 55.1 97.5 91.6 70.7 97.45 83.0 44.7 98.3 91.7 73.1 99.16 85.9 53.1 97.2 92.7 77.5 97.87 86.4 62.0 96.9 93.2 83.4 97.58 86.1 52.7 97.6 91.6 72.7 98.1

Abbreviations: IMRT � intensity-modulated radiotherapy; CT �omputed tomography; CTV � clinical target volume; SV �eminal vesicle.

respectively. The dotted line shows the results using IMRT.

We observed that the CTV was not adequately coveredor some patients (e.g., Patient 9) even when the prostateenter of volume method was used. The main reason is thathe CTV for this patient included both the prostate and theeminal vesicles. The CT-guided method proposed in thistudy uses only the prostate as the reference for setting upreatments. The seminal vesicle sometimes did not moveith the bulk of the prostate. Table 3 shows the CTV,rostate, and seminal vesicle volumes receiving the pre-cribed dose (75.6 Gy) for the eight daily CT plans for bothhe proton therapy and IMRT plans for Patient 9. Therostate was adequately covered by both the IMRT androton therapy plans, but the seminal vesicles were under-osed. For this patient, the 97% of the seminal vesseleceived at least 60 Gy for both IMRT and proton plans.owever, even in this worst case, the proton therapy planas less sensitive to interfractional variation than the IMRTlan was.Although comparing treatment planning for IMRT and

roton therapy was not a main focus of this study, we

modulated radiation therapy (IMRT) plan (a), a protonopposed beams, (c) the comparison of dose–volumeand bladder in the prostate cancer case. The solid androton beams and two parallel opposed lateral beams,

ensity-arallelectum,lique p

oro(yAuwuttbphRftsbw

fotbb

t(haottppa

oHsdtgb

cfn

1

1


bserved that the IMRT plans were better at sparing theectum at doses higher than 50 Gy for most patients. Webserved that if we only treated partial seminal vesselsproximal seminal vessels), proton plan would normallyield better rectum sparing at doses higher than 50 Gy.nother reason for this is that only two lateral beams weresed for the proton treatment plans. To study this further,e designed another proton plan for one of the patientssing four beams—two lateral, parallel opposed beams andwo oblique beams. Figure 7 shows the dose distributions inransverse planes for an IMRT plan (Panel a), the protoneam with four beams (Panel b), the proton plan with twoarallel-opposed beams (Panel c), and the dose–volumeistograms (Panel d) for the PTV, rectum, and bladder.ectum sparing for doses higher than 50 Gy is very similar

or IMRT and the four-beam proton plans. For doses lesshan 50 Gy, the four-beam proton plan is better at rectumparing than the IMRT plan. We also observed that theetter sparing of the rectum by the four-beam proton planas achieved at the expense of bladder sparing.The main reason why most proton centers do not use such

our-beam or similar multibeam approaches is that theblique beams would be aimed at the rectum. For protonherapy, there is a substantial uncertainty as to where theeam actually stops. In addition, there is uncertainty in
iologic equivalence of the physical proton dose compared m
REFEREN

ison of daily prostate alignment utilizing skin marks, ultra-

1

1

1

1

1

1

1

1

2

2

o the same photon dose. The radiobiological effectivenessRBE) of proton dose known to increase with depth and isigher for lower doses. Such variations in RBE are ignorednd a fixed value of 1.1 is used. To avoid the consequencesf range uncertainty and RBE approximations, a rule ofhumb in proton treatment planning is adopted, which is noto aim the proton beam toward the critical organs in theroximity. We believe that the quality of proton therapylanning will improve significantly if uncertainties in rangend RBE are reduced.

CONCLUSION

Interfractional variations in volumes, positions, and shapesf targets and critical normal tissues can be significant.owever, for proton therapy plans designed using the pas-

ive scattering technique, changes in the dose distributionue to interfractional anatomic changes were no worse thanhose for IMRT plans when adequate consideration wasiven to the additional uncertainties caused by protoneams.We believe that there is considerable potential for better

ontrol of the range uncertainties for proton beams andurther improvement of proton treatment planning tech-iques, which will be needed for intensity- and energy-
odulated proton therapy.
CES

1. Smith AR, Intensity-modulated conformal radiation therapyand 3-dimensional treatment planning will significantly re-duce the need for therapeutic approaches with particles such asprotons—Against the proposition. Med Phys 1999;26:1187.

2. Goitein M, Lomax AJ, Pedroni ES. Treating cancer withprotons. Phys Today 2002;55:45–50.

3. Jones B, Burnet N. Radiotherapy for the future—Protons andions hold much promise. BMJ 2005;330:979B–980B.

4. Goitein M, Lomax AJ, Pedroni ES. Weighing proton therapy’sclinical readiness and costs. Phys Today 2003;56:13–14.

5. Wong JR, Grimm L, Uematsu M, et al. Image-guided radio-therapy for prostate cancer by CT-linear accelerator combina-tion: Prostate movements and dosimetric considerations. Int JRadiat Oncol Biol Phys 2005;61:561–569.

6. Mohan R, Zhang XD, Wang H, et al. Use of deformedintensity distributions for on-line modification of image-guided IMRT to account for interfractional anatomic changes.Int J Radiat Oncol Biol Phys 2005;61:1258–1266.

7. Yan D, Vicini F, Wong J, et al. Adaptive radiation therapy.Phys Med Biol 1997;42:123–132.

8. Yan D, Ziaja E, Jaffray D, et al. The use of adaptive radiationtherapy to reduce setup error: A prospective clinical study. IntJ Radiat Oncol Biol Phys 1998;41:715–720.

9. Yan D, Xu B, Lockman D, et al. The influence of interpatientand intrapatient rectum variation on external beam treatmentof prostate cancer. Int J Radiat Oncol Biol Phys 2001;51:1111–1119.

0. Court LE, Dong L, Taylor N, et al. Evaluation of a contour-alignment technique for CT-guided prostate radiotherapy: Anintra- and interobserver study. Int J Radiat Oncol Biol Phys2004;59:412–418.

1. O’Daniel JC, Dong L, Bonnen M, et al. Dosimetric compar-

sound, and in-room CT. Int J Radiat Oncol Biol Phys 2004;60:S617–S618.

2. Rehbinder D, Forsgren J, Löf J. Adaptive radiation therapy forcompensation of errors in patient setup and treatment delivery.Med Phys 2004;31:3363–3371.

3. Yan D, Wong J, Vicini F, et al. Adaptive modification oftreatment planning to minimize the deleterious effects of treat-ment setup errors. Int J Radiat Oncol Biol Phys 1997;38:197–206.

4. Yan D, Lockman D. Organ/patient geometric variation inexternal beam radiotherapy and its effects. Med Phys 2001;28:593–602.

5. Mackie TR. Intensity-modulated conformal radiation therapyand 3-dimensional treatment planning will significantly re-duce the need for therapeutic approaches with particles such asprotons—For the proposition. Med Phys 1999;26:1185–1186.

6. Slater JD, Yonemoto LT, Rossi CJ, et al. Conformal protontherapy for prostate carcinoma. Int J Radiat Oncol Biol Phys1998;42:299–304.

7. Slater JD, Rossi CJ, Yonemoto LT, et al. Proton therapy forprostate cancer: The initial Loma Linda University experi-ence. Int J Radiat Oncol Biol Phys 2004;59:348–352.

8. Moyers MF, Miller DW, Bush DA, et al. Methodologies andtools for proton beam design for lung tumors. Int J RadiatOncol Biol Phys 2001;49:1429–1438.

9. Moyers MF, Miller DW. Range, range modulation, and fieldradius requirements for proton therapy of prostate cancer.Technol Cancer Res Treatment 2003;2:445–447.

0. Urie M, Goitein M, Wagner M. Compensating for heteroge-neities in proton radiation-therapy. Phys Med Biol 1984;29:553–566.

1. Thomas SJ. Margins for treatment planning of proton therapy.
Phys Med Biol 2006;51:1491–1501.

effect of anatomic motion on proton therapy dose distributions in prostate cancer treatment

Documents