Portal Imaging Protocol for Radical Dose-Escalated Radiotherapy Treatment of Prostate Cancer

PII S0360-3016(97)00551-8

● Physics Contribution

PORTAL IMAGING PROTOCOL FOR RADICAL DOSE-ESCALATEDRADIOTHERAPY TREATMENT OF PROSTATE CANCER

C. D. MUBATA, PH.D.,* A. M. BIDMEAD, M.SC.,* L. M. ELLINGHAM , M.SC.,* V. THOMPSON†

AND

D. P. DEARNALEY, M.D.†

*The Joint Department of Physics. The Royal Marsden NHS Trust and The Institute of Cancer Research, Fulham Rd., London SW3 6JJ, UK;and†Department of Radiotherapy, The Royal Marsden NHS Trust and The Institute of Cancer Research, Fulham Rd., London SW3 6JJ, UK

Purpose: The use of escalated radiation doses to improve local control in conformal radiotherapy of prostatic canceris becoming the focus of many centers. There are, however, increased side effects associated with increased radio-therapy doses that are believed to be dependent on the volume of normal tissue irradiated. For this reason, accuratepatient positioning, CT planning with 3D reconstruction of volumes of interest, clear definition of treatment marginsand verification of treatment fields are necessary components of the quality control for these procedures. In this studyelectronic portal images are used to (a) evaluate the magnitude and effect of the setup errors encountered in patientpositioning techniques, and (b) verify the multileaf collimator (MLC) field patterns for each of the treatment fields.Methods and Materials: The Phase I volume, with a planning target volume (PTV) composed of the gross tumourvolume (GTV) plus a 1.5 cm margin is treated conformally with a three-field plan (usually an anterior field andtwo lateral or oblique fields). A Phase II, with no margin around the GTV, is treated using two lateral and fouroblique fields. Portal images are acquired and compared to digitally reconstructed radiographs (DRR) and/orsimulator films during Phase I to assess the systematic (CT planning or simulator to treatment error) and thedaily random errors. The match results from these images are used to correct for the systematic errors, ifnecessary, and to monitor the time trends and effectiveness of patient imobilization systems used during thePhase I treatment course. For the Phase II, portal images of an anterior and lateral field (larger than thetreatment fields) matched to DRRs (or simulator images) are used to verify the isocenter position 1 week beforestart of Phase II. The Portal images are acquired for all the treatment fields on the first day to verify the MLCfield patterns and archived for records. The final distribution of the setup errors was used to calculate modifieddose–volume histograms (DVHs). This procedure was carried out on 36 prostate cancer patients, 12 withvacuum-molded (VacFix) bags for immobilization and 24 with no immobilization.Results: The systematic errors can be visualized and corrected for before the doses are increased above theconventional levels. The requirement for correction of these errors (e.g., 2.5 mm AP shift) was demonstrated,using DVHs, in the observed 10% increase in rectal volume receiving at least 60 Gy. The random (daily) errorsobserved showed the need for patient fixation devices when treating with reduced margins. The percentage offields with displacements of<5.0 mm increased from 82 to 96% with the use of VacFix bags. The rotation of thepelvis is also minimized when the bags are used, with over 95% of the fields with rotations of<2.0° comparedto 85% without. Currently, a combination of VacFix and thermoplastic casts is being investigated.Conclusion: The systematic errors can easily be identified and corrected for in the early stages of the Phase I treatmentcourse. The time trends observed during the course of Phase I in conjunction with the isocenter verification at the startof Phase II give good prediction of the accuracy of the setup during Phase II, where visibility of identifiable structuresis reduced in the small fields. The acquisition and inspection of the portal images for the small Phase II fields has beenfound to be an effective way of keeping a record of the MLC field patterns used. Incorporation of the distribution ofthe setup errors into the planning system also gives a clearer picture of how the prescribed dose was delivered. Thisinformation can be useful in dose–escalation studies in determining the relationship between the local control ormorbidity rates and prescribed dose. © 1998 Elsevier Science Inc.

Portal images, Treatment setup errors, Dose escalation, Treatment planning.

INTRODUCTION

The improvement of techniques in treatment of prostatecancer has become the objective of many centers. The use of

blood markers such as Prostate-Specific Antigen (PSA) hasled to the early detection of these cancers at a fairly local-ized and potentially curable stage. Radical radiotherapy ismostly used as one of the curative treatments. In a number

Reprint requests to: Cephas D. Mubata, Ph.D., The Joint De-partment of Physics. The Royal Marsden NHS Trust and TheInstitute of Cancer Research, Fulham Rd. London SW3 6JJ, UK.Acknowledgments—This project was partly sponsored by VarianOncology Systems (Crawley, UK). The authors would like to

thank the radiotherapy radiographers who helped in the acquisitionand matching of the portal images, and also some useful feedbackin the development of the protocol.

Accepted for publication 22 July 1997.

Int. J. Radiation Oncology Biol. Phys., Vol. 40, No. 1, pp. 221–231, 1998Copyright © 1998 Elsevier Science Inc.Printed in the USA. All rights reserved

0360-3016/98 $19.001 .00

221

of centers (8, 18, 24) including ours, dose-escalated radio-therapy studies are being performed for treatment of pros-tate cancer using conformal 3D treatment planning to im-prove local control without increasing radiation inducedside effects. Although the increase of radiotherapy dose isthought to improve local control, local failure is more prom-inent with larger tumors. The use of hormone therapy suchas androgen deprivation has been reported to reduce thevolume of the prostate and, therefore, potentially the treat-ment target volume by up to 50 and 40%, respectively (6,25, 35). This reduction of the treatment volume can enablethe doses to be escalated to doses above conventional levelswithout increased normal tissue complications that are de-pendent on the amount of normal tissue irradiated. Theamount of normal tissue treated has been reported to bereduced by 41% [with 40 and 42% reductions in rectal andbladder volumes irradiated, respectively (7, 26)].

At the Royal Marsden, a randomized trial of dose esca-lation using radical conformal radiotherapy for localizedprostate cancer following readjuvant androgen deprivationis currently being performed. The main aims of the dose-escalation project are to determine the effect of increasedradiation dose on (a) local control, (b) radiation side effects,and (c) the evaluation of optimal treatment volumes afterneoadjuvant androgen deprivation. Electronic portal imag-ing is used to evaluate the accuracy of patient positioningduring treatment and to provide information on how thetreatment course was delivered. The latter is achieved byinvestigating the effect of the setup errors on the dose–volume histograms (DVH) and the resultant tumor controlprobabilities (TCP) and normal tissue complication proba-bilities (NTCP) calculated from the original treatment plan.The portal imaging protocol employed to verify and correctpatient positioning will be presented.

METHODS AND MATERIALS

Treatment planningThis study was carried out on 24 patients with Stage

T1b–T3 prostate cancer. The patients are classified into twogroups: group 1 with low risk of seminal vesicle involve-ment (clinical Stage T1b–T2b), and group 2, moderate orhigh risk of seminal vesicles involvement with T2c or T3staging. For the first treatment phase (Phase I), for group 1,the gross tumor volume GTV is defined as the prostate andbase of seminal vesicles and prostate plus seminal vesiclesfor group 2. A standardized GTV of prostate and base ofseminal vesicles is used for all patients during the Phase II.

Patients are additionally randomized between standardvolume and reduced volume after hormone therapy. Patientsin the reduced volume arm are treated with a (CTV) clinicaltarget volume of GTV1 0.5 cm, while in the standardvolume the CTV is composed of a volume derived from theoriginal extent of the disease before hormone therapy (6)plus a 0.5-cm margin added to the GTV. The patients arescanned supine with a full bladder using 0.5-cm slices takenfrom 1.0 cm inferior to the ischial tuberosities to the domeof the bladder. Outside this volume scans are taken at1.0-cm spacing. The treatment planning was originally car-ried out on Target II (IGE) and later on the new CADPLAN(Varian Oncology systems) planning systems. The latterwas preferred immediately after commissioning because ofthe better MLC interface and more advanced planning toolsincorporated on the planning systems. For the Phase I, thePTV, composed of the CTV plus a 0.5-cm margin, is treatedconformally with a three-field plan (usually an anterior fieldand two lateral or oblique fields). The Phase II is treatedwith six coplanar conformal fields with two laterals, two an-terior obliques, and two posterior oblique fields (see Fig. 1).

Fig. 1. Typical three-field Phase I (left) and six-field Phase II (right) plans used in the dose-escalated conformalradiotherapy of prostate cancer.

222 I. J. Radiation Oncology● Biology ● Physics Volume 40, Number 1, 1998

All fields are conformally shielded using multileaf collima-tors. The MLC leaves are fitted to the PTV with a 6.0 mmmargin (to account for beam penumbra) in each beam’s-eye-view (BEV). The equal area or transection conventionof leaf fitted was used. With no MLC interface on the TargetII system, the BEV images were printed out, a 6.0 mmmargin added, and the shielding contour then digitised onthe MLC Shaper™ system (Varian Oncology Systems). TheMLC interface on CADPLAN enables automatic creationand direct transfer of the leaf patterns to the MLC worksta-tion on the Clinac. The dose prescription to the standarddose patients is 64 Gy in 2 Gy fractions, with an additional10 Gy in five fractions given to the boost volume (Phase II)for the patients in the dose-escalated arm of the trial. Ap-propriate simulator films are taken to verify the orientationand alignment of the fields and isocenter position.

Treatment setupOf the 36 patients included in this study, 12 patients were

located in vacuum-molded (VacFix) bags, with ankle sup-ports. The VacFix bags were located on a special styrofoamboard that keys into the top of the couch (see Fig. 2). Theboard serves a dual purpose, first to locate the VacFix bags(see Fig. 2b), and second, to enable portal images to beacquired for all fields without the interference of couch bars.The Clinac couch top is designed such that there is a centralspine at one end of the couch that is designed for clearanceof postoblique beams. The AP section of the couch hasmetal side bars that restricts the use of oblique fields on thissection. It becomes difficult, therefore, to acquire ant- andpostoblique fields at the same treatment fraction withouthaving to turn the couch around. The acquisition of anteriorand oblique or lateral fields together is essential to interpretsome of the out of plane rotations that are difficult tovisualize with 2D portal images in one plane. This boardallows acquisition of all the images resulting in more accu-rate analysis of these 2D images.

Reference imagesThe image matching process starts with the input of the

reference image, usually the simulator film or DRRs. Whenthe simulator film is used as the reference image, the fieldedges are clearly highlighted and the MLC leaf patterns areclearly marked, then the film was digitized using a CCDcamera. An example of the simulator image showing fieldedge in blue is shown in Fig. 3. The DRRs are generated onthe treatment-planning system and imported from the Com-puter Aided Radiotherapy (CART) format to PortalVi-sion™ image format. Software has been written that facil-itates scaling and calculation of the magnification on thereference image by engraving two points, which are aknown distance apart, onto the image data. The coordinatesof the field corners, MLC leaf positions, and the coordinatesof any structure of interest (if required) are transferreddirectly to the electronic portal imaging (EPI) system data-base in binary form compatible with the match structurefiles in the EPI database (Fig. 3) (32). The field edge

parameters are used in the calculation of field displacementsduring image matching, while the contour of the otherstructures transferred from the planning system, are used forvisualisation purposes only. After the reference image has

Fig. 2. Photographs of the styrofoam board that is keyed securelyonto the couch top to provide clearance from the couch side metalbars for the posterior oblique fields and to locate the VacFix bags.The top of the board showing the VacFix bag location point isshown in b.

223Portal imaging protocol for prostate cancer● C. D. MUBATA et al.

been acquired, a once only preparation of the referenceimage is carried out. This involves defining the field edgesand at least three match structures (preferably fixed bonyanatomical structures) used in matching of the portal imagesto the reference image (9, 10, 32). The structures (includingthe defined field edge) are displayed as graphs overlaid onthe images.

Portal imagesThe portal images were acquired using a matrix ionisa-

tion detector (Varian PortalVision™) attached to Clinac2100C with a retractable (robotic) arm. For the Phase I, theportal images are acquired during treatment delivery usingthe treatment fields. If the image quality is not adequate,which sometimes occurs especially for lateral fields, thendouble-exposure images are acquired on the following day.For the double-exposure images, the planned or treatmentfield images are acquired during treatment fraction deliveryand the open field acquired immediately after with a largerthan treatment field (163 16 cm for AP and 163 14 cmlateral fields) at a reduced dose rate of 80 monitor units perminute (MU/min). At this dose rate about 8 MU are deliv-ered during the acquisition of the pen field. The detectortakes about 5.3 s to acquire an image in the standardacquisition mode. At a dose rate of 400 MU/min the imageis acquired in about 37 MU. The planned and open fieldimages are combined as shown in Fig. 4. Daily acquisitionof the portal images is continued until the setup has beenapproved then weekly images are acquired during the rest ofthe treatment course. In this study the number of dailyimages acquired for each patient ranged from three to eight,depending on the day the treatment commenced in relation

to the clinical audit day, also on the accuracy of the setup.Of the 36 patients included in this study, 3 required aresimulation and 2 had their setups changed on set, and acontinuation of daily imaging until the second audit meet-ing. There were no second corrections needed for all thepatients included in this group. For the Phase II isocentercheck, double-exposure images are acquired for two orthog-onal fields (anterior and lateral) a week before the patient isdue to start the Phase II. On the first day of treatment, theimages of all the six conformal fields are acquired and keptas a record of the MLC fields used. The portal imagesacquired using the boost treatment fields are too small, anddo not contain enough identifiable anatomical structures forreliable image matching to be carried out.

Image analysisThe matching of the portal to the reference image is

carried out interactively. In cases where the defined refer-ence structures are well delineated, an automated computer-assisted match can be carried out using the feature extrac-tion algorithms of Gilhuijs (9, 11). This is typical forAP–PA pelvis fields where the edges of the pubic bones andthe obturator foramen are well differentiated. For lateral oroblique pelvis fields, the need arises for observer’s inter-vention in the matching process to manually align the linegraphs of structures from the reference image, with thevisible structures on the portal images.

The match data obtained throughout the treatment coursewere used to recalculate the DVHs for the different iso-center shifts. For the patients in the standard dose arm, theDVH recalculations were carried out for the three-field plan.

Fig. 3. Typical simulator and DRR reference images used in image matching. The simulator image shows the MLC fieldedge and the edges of bony anatomy used as reference structures. The DRR shows some of the contours transferred fromthe planning systems, such as the GTV, PTV and femoral heads together with the field edges.


The composite nine (three Phase I and six Phase II) fieldplan was used for the escalated dose arm calculations.

RESULTS

Setup deviationsThe patient positioning errors can be separated into the

systematic (simulator treatment) and the random (daily)error. The typical distribution of the systematic and random

errors observed in these, conformal fields is shown in Fig. 5.The distributions of the systematic and random errors aresimilar, as observed also by Rabinowitz (23), with a stan-dard deviation of 2.5 mm. Our previous studies (20, 31)have shown that both the systematic and random errors aremore frequent in the AP direction than lateral and sup–inf.directions. The lateral tattoo marks are not as reliable as theanterior marks where the symphasis pubis can easily belocated. The lack of firm bony structures which can be

Fig. 4. Orthogonal simulator and the corresponding portal images used in isocenter verification using double exposuretechniques. The larger open field is used for the calculation of the field displacement while the smaller (treatment) fieldis used to provide a visual effect of the displacement on the treatment area.


‘‘felt’’ on the sides of the patient during simulation, and thestretching of the skin reduces the reliability of the locationof the isocenter height relative to these exterior marks.These errors can be reduced by the use of extra isocenterchecks such as maintaining fixed couch and tattoo heights.

Verification and correction procedureThe systematic error needs to be identified and corrected

for as soon as possible during the first few fractions. Theaverage of the translational and rotational errors of the firstfew (approximately five) images was found to be a goodestimate of the average over the whole treatment course. Afirst-hand analysis of images is carried out by the radiogra-phers. This analysis is carried out off-line and if any grosserrors (.6.0 mm) are observed then the setup is reviewedwith a view to adjust the next setup. The results of theanalysis of approximately five images is presented atweekly clinical planning audit meetings. The results arenormally presented in the form of a reference image withthe field edges of the subsequent portal images superim-posed on it, as shown in Fig. 6. The actual average shifts arequantified for all the fields. If there is a significant system-atic error in the setup, then the decision to take correctivemeasures is taken by the clinician at these meetings.

There are no set intervention limits used for correction ofsystematic setup errors. Each setup is reviewed on its ownmerit. If DRRs are used, the effect of systematic shift isevaluated more clearly by taking into account the positionof the target volume and the contours of the surroundingcritical organs transferred from the planning system (seeFig. 6c). The corrective action depends on the referenceimage used. If a simulator image is used as a referenceimage, then the treatment area location is reviewed on thesimulator and portal images. The simulator images maysuffer from random transfer errors (2). It cannot, therefore,be assumed that the simulator is the ‘‘gold standard.’’ On

Fig. 6. The reference images with the field edges from consecutiveportal images superimposed for (a) the anterior and (b) the lateralfields. The contour of the PTV can be displayed alongside theportal image edges when DRR’s (c) are used.

Fig. 5. Comparison of the random and systematic errors in the APdirection encountered in the lateral and oblique conformal pelvicfields, for all the patients with and without VacFix bags.


reviewing the simulator and portal images the clinician candecide whether to change the setup on-set or request aresimulation. The DRRs derived from tomographic dataused for planning are assumed to be more correlated withthe expected portal images. If a systematic shift needs to becorrected, the setup can be changed on-set and a new set ofskin marks introduced. Three out of 36 patients were re-simulated with systematic errors of 3.0–4.2 mm. Afterrematching of the images to new reference images wascarried out, no further corrections were necessary after thefirst audit meeting for any of these cases. A change in thesetup mark position was also carried out on-set for two morepatients after systematic errors of 2.5 and 3.4 mm werenoticed. The use of portal image edge plots superimposedon the reference images (see Fig. 6), tends to highlight thesystematic error, so that one is obligated to correct for evensmall discrepancies. Sometimes the clinician can feel com-peled (just from the visual display of the results) to correctfor setup that in principle have been accounted for in thetreatment margins! Once the setup has been approved, thenweekly images are acquired to monitor time trends (3).Typical scatter plots used to display the results over a periodof time in monitoring time trends is shown in Fig. 7. Thegradual shift due to changes in patient separation can easilybe identified. The time trend information is also useful inassessing the overall accuracy or reproducibility of thepositioning of the patient before the doses are escalated tovalues above the conventional levels in the Phase II.

A week before the Phase II commences, two orthogonaldouble exposure images are acquired, centered on the PhaseII isocenter, and matched to the set of verification filmsacquired in the simulator or DRR. This check is carried outa week before Phase II begins, so that any discrepancies can

be addressed before the additional dose is given to thereduced volume. The portal images acquired using the smallconformal fields are not very useful for image matching dueto lack of identifiable match structures. These images areonly taken on the first day of Phase II and used as a recordof the actual MLC fields used.

Patient fixationAlthough it is straightforward to correct for the system-

atic errors, the only way to correct for the random errors isto use on-line verification and correction procedures (5, 14).Ten Hakenet al. (30) has developed a tilt and roll treatmenttable for automatic correction of setup errors. This does nottake in account the variation of position and shapes of theinternal organs, as the orientation of the table top changes.The use of different fixation devices during radiotherapy ofthe prostate has been reported by different groups (18, 27).These include the alpha cradle, aquaplast (27), polycast (13,27) and vacuum-molded bags (20). These immobilizationdevices have been used by these groups with varying de-grees of success. We have used vacuum-molded bags as afixation device in attempting to reduce the random error.Contrary to the findings of Songet al. (27), the VacFix bagshave been found to improve the setups significantly as canbe observed in Fig. 8. Not only do the bags reduce thetranslational shifts but also help to maintain the orientationof the pelvis. Figure 9 shows the difference in the distribu-tion of rotational errors about the right–left (R–L) axis forpatients with and without VacFix bags. When these bags areused, 95% of the images analyzed showed rotations of,2.0° for the lateral and oblique fields. Without the bags,the percentage drops down to 85%. It is important to curbthe rotation about R–L axis because this motion can signif-icantly affect the volume of the prostate, bladder and rectumirradiated (33). With the VacFix bags our studies haveshown that the sides of the bags have to be kept low. Thishelps the patient to get into the mold without overstretching

Fig. 7. Scatter plots of the daily setup variations showing the timetrends of the patient positioning over a period of time.

Fig. 8. The effect of vacuum-molded bags on distributions of thetranslational setup errors in sup–inf and AP directions, observed inlateral and oblique fields.


the skin. The low sides also enable easier adjustment of thepatient’s position or ‘‘tilt’’ by the radiographers if the marksare not aligned properly. The skin marks on the patient arealso visible and not covered by high sides. The length of thebags also have to be long enough so that the contour of thepatient’s back can be followed over a reasonable length (70cm long is adequate, without being too large).

Dose–volume histograms (DVHs)The effects of the patient setup variations errors on the

dosimetry were investigated by calculating the DVHs basedon the matched data from the portal images. With thesystematic errors corrected for using the verification andcorrection procedure described above, only the random er-rors were considered in this exercise. Examples of thevariations of the DVHs for the GTV, bladder, and rectumare shown in Fig. 10. These histograms were calculated withthe isocentre shifted by distances equal to the standarddeviation of the systematic errors in the AP and PA direc-tions. The setup errors in the AP direction are shown herebecause they are the most prominent (20) and have the mostsignificant effect on the rectal and bladder doses. For ex-ample, a shift of 2.5 mm in the AP direction can increase thevolume of the rectal wall receiving 60 Gy by about 10%.These DVHs highlight the need for implementation of ver-ification and correction procedures to correct for systematicerrors, in these dose-escalated treatment setups, to reducecommon complications such as rectal bleeding (4).

An analysis of the GTV (or PTV) histograms recalculatedfrom the match data is useful in the evaluation of theeffectiveness of the reduced margins on the coverage ofthese volumes. A typical example is shown in Fig. 10a, in acase where the GTV is given a boost of 10 Gy Phase II withno margin (PTV5 GTV). In this case the volume of theGTV receiving 74 Gy could be reduced by up to 15%, withthese relatively small translational displacements of theisocentre. The DVHs in Fig. 10 represent just a single caseand the volumes vary from patient to patient and with

treatment parameters such as margins and prescription dose.A more detailed comparison of the DVHs for the differentpatients, together with combinations dose and volume arms,is part of an on going study which will not be discussedhere.

DISCUSSION

With the dose–escalation studies, the main goal is toestablish the relationship between the dose delivered andsuccess or failure rates. Monitoring of the patient position-ing during treatment is, therefore, necessary to ensure min-imum variation in the irradiated volume. With critical struc-tures, especially the rectum located in the vicinity of theprostate, there is a need to correct for the systematic errorsas soon as possible at the beginning of the treatment course.The portal images acquired during the Phase I course,together with the double exposure images have been found

Fig. 9. The effect of the vacuum-molded bags on the pelvisrotation about the right–left axis.

Fig. 10. The cumulative dose–volume histograms for the GTV andrectum calculated for the original plan (solid), and with the iso-center shifted one standard deviation of the systematic errors (2.5mm). AP (dot-dash) and PA (dotted), for a prostate plan receiving74 Gy to a standard volume.


to be adequate to identify these systematic errors. The PhaseII portal images are too small and do not contain identifiableanatomical structures that can be used for image matching,if only the treatment field images are used. Monitoring ofthe time trends from the Phase I can provide information onthe reproducibility of the setup. The only information on theaccuracy of the treatment setups during Phase II is, there-fore, the double-exposure verification check carried outtowards the end of the Phase I and the analysis of the matchdata from the Phase I. It is possible to acquire double-exposure (or just single larger-than-treatment field) imagesusing two orthogonal (one AP and one lateral) during thePhase II session. Considering that the Phase II is only fivefractions long, and at this stage the image acquisition fre-quency will be on weekly, there is no need to increase thisfrequency of imaging if the time trends do not show anyincrease in the magnitude of the random errors. The choiceof whether to acquire additional larger-than-treatment fieldimages for these last five fractions depends on individualdepartment policies and time constraints. Another way ofmonitoring the patient positioning during dose delivery tothe boost volume is to use the simultaneous boost tech-niques of Lebesque and Keus (16). With these techniquesthe anatomical structures from the original larger volumecan be used for image matching.

The need to correct for the systematic errors has beendemonstrated using the DVHs. These systematic errors canbe corrected using either on-line portal imaging or by theuse of verification and correction procedures. There aresome correction procedures that have been used by othergroups (3, 5). The shrinking margins used by the Nether-lands Cancer Institute (34) has been reported to be verysuccessful in correction of these errors. The process isautomated to take the burden off the clinician, but it is notquite clear what plan of action is taken once the systematicerror has been identified as above the limits. The DVHs ofthe PTV and GTV can be used in the evaluation of theeffectiveness of the clinical margins applied in these tech-niques. This is especially useful in these studies because oneof the aims of the dose–escalation studies is actually todetermine the margins to be used for these treatments afterhormone therapy. This information has to be augmentedwith some form of monitoring of the movements of internalstructures. The feasibility of using markers implanted ontothe periphery of the prostate has been reported by Balter (1)and is also being investigated at our institute. Lebesqueetal. (17) used repeat CT scans and have shown that theprostate can be displaced by up to 1.0 cm due to rectal orbladder filling. This magnitude of shift has also been re-ported by Ten Haken (29). In our case, the only informationavailable about the internal movements of the prostate isfrom measurements carried out using dynamic magneticresonance imaging (MRI) (22). With these measurementsrepeat scans are acquired over a period of 10 min and theshape and position of the prostate monitored during thisperiod. This gives information about the possible move-ments of the target volume over that short space of time.

Because this is carried out before commencement of theradiotherapy treatment course, it does not provide informa-tion about the movement of the prostate throughout the6-week duration of the course. The elasticity of the organwall changes due to radiation damage during the course ofradiotherapy leading to variations in the effects of rectal orbladder filling as reported by Lebesque (17). This techniquedoes not, however, provide information at the time of treat-ment fraction delivery, so cannot be correlated to the portalimage information. A way of combining the portal imageinformation and internal organ movements would give aclearer picture in margin evaluations and dose responsederivatives.

In their studies using repeat CT scans, van Herket al.(33) reported that there is a strong correlation between theorientation of the pelvis and the shape and position of theprostate. The rotation of the pelvis about the right–left axiswas found to have the biggest effect. There is a need,therefore, to use molding mattresses such as the alpha cradleor the vacuum-molded bags. In the implementation of thesebags, investigations have to be carried out to find the opti-mum size, depending on the treatment techniques. Theoptimum depths and lengths of such devices have to beevaluated in terms of ease of use by the radiographers,relocatability of the patients within the support and patientcomfort. Our low-sided VacFix bags have been found to beeffective in reducing the rotations of the pelvis about theR–L axis. The use of polycast shells to immobilize thepelvis tends to be uncomfortable for a patient lying supinewith a full bladder. The ankle supports are also useful inmonitoring the leg orientation, which also to a lesser extentaffect the position of the target volume (33).

Incorporation of patient setup errors in treatment plan-ning systems has been reported by other institutions (12, 13,19). With the systematic errors corrected for and the randomerrors following a normal distribution with a mean of zero,the interpretation of the effect of these on the resultantdistribution is a nontrivial problem. It is a cumbersomeprocess to manually simulate all the daily isocentre dis-placements obtained from the portal images. This could befacilitated by incorporating algorithms within the planningsystem to utilize such data in an automated way as carriedout by Mageras (19). In their work Mageraset al. usedinformation on organ motion obtained from repeat CT scansof a group of 12 patients, as reference data used to simulatethe probable motions during treatment planning. This sam-ple of 12 patients was assumed to be representative of thewhole population. So, for each patient the organ contourswere adjusted based on the variations obtained from thereference patients. A much larger database of referencepatients is required to account for differences in patientsorientation, rectal and bladder filling, and the effects ofradiation to the internal organs, for improved statisticalaccuracy. Kutcheret al. (15) have suggested the use ofaverage DVHs together with the 10 and 90% confidencelimit on dose–volume histograms (CLDVHs). If one obtainsthe DVHs for use in TCP and NTCP calculations, then the


question which arises is ‘‘which DVH to use?’’ for thesecalculations. Usually the DVHs from the original plan areused to arrive at a single TCP value. The TCP calculationsare based on the clonogenic cell killing in an organ orvolume receiving a homogeneous prescribed dose in equalfractions over a period of time. The actual meaning of theTCPs calculated for these GTV receiving a nonhomoge-neous dose still needs to be addressed. Niermierko andGoitein (21) have proposed a model that uses biologicalnormalised DVHs for estimating the TCP for an inhomo-geneously irradiated tumor. Their model accounts for situ-ations were the dose distribution in a tumor is nonuniformsuch that part of the volume is underdosed by, say, 10% ofthe prescribed dose. The model does not account for theday-to-day mobility of the tumor volume because a constantdose per fraction is assumed to be delivered in a knownnumber of equal fractions. More recently, Stavrevet al.(29)have proposed the use of a modified or reduced cell densityin the region where the tumor volume drifts in and out of theintended or nonmobile state. Using Monte Carlo simula-tions, they obtained TCP calculations that were independentof the variance of the mobility of the target volume becausethese movements were accounted for in the dose prescrip-tion. It is, however, not explained how this effect the dosesto the neighboring normal tissue (i.e., the effect on NTCP ofsurrounding organs). Moreover, their calculations assume arigid nondeformable target volume that does not hold for theprostate gland due to rectal and bladder filling together withthe rotation of the pelvis about the R–L axis.

We have used the distributions of the setup errors ob-tained for each patient to calculate probability distributionsof organ volume receiving a specified dose. For example,Fig. 11 shows the probability distribution of rectal volumesreceiving a specified dose. For example, Fig. 11 shows theprobability distribution of rectal volumes receiving a TD 5/5(tolerance dose with 5% probability of complication within5 years) of 60 Gy. More information could be extractedfrom such distributions to augment the single TCP values,calculated from average or intended plans.

CONCLUSION

A simple verification and correction procedure derived tocorrect for systematic simulator-treatment field deviations

during Phase I has been presented. The systematic errorsthat need to be corrected as soon as possible can be deter-mined in the first few daily portal images and corrected forif need be. After correction of these errors weekly imagescan be acquired for the rest of the Phase I course. The timetrends observed during Phase I in conjunction with theisocenter verification before the start of Phase II have beenfound to be adequate for setup verification during Phase II,where visibility of identifiable structures is reduced in thesmall fields. The acquisition and analysis of the portalimages for these small fields has been found to be aneffective way of checking and keeping a record of MLCfield patterns used during treatment. The use of fixationdevices in these dose escalated trials is also recommendedin order to minimize the random translational and rotationalpatient positioning errors. Incorporation of the distributionof the setup errors into the planning system also gives aclearer picture of how the prescribed dose was delivered.The effect of the these setup deviations on the calculation ofthe TCPs and NTCPs still needs further investigation, to-gether with monitoring of internal organ movements. Thisinformation could be useful in dose–escalation studies fordetermining the relationship between the local control ormorbidity rates and prescribed dose.

REFERENCES

1. Balter, J.; Lam, K.; Sandler, H. F.; Littles, F.; Ten Haken, R.Automatic localization of the prostate at the start of the treat-ment using implanted radio-opaque markers (Abstract). In:Proceedings of the third international workshop on portalimaging. San Francisco: Manitoba Cancer Treatment and Re-search Foundation; 1994.

2. Bel, A.; Bartelink, H.; Vijlbrief, R. E.; Lebesque, J. V. Trans-fer errors of planning CT to simulator: A possible source ofsetup inaccuracies? Radiother. Oncol. 31:176–180; 1994.

3. Bel, A.; van Herk, M.; Bartelink, H.; Lebesque, J. V. A

verification procedure to improve patient set-up accuracy us-ing portal images. Radiother. Oncol. 29:253–260; 1993.

4. Benk, V. A.; Adams, J. A.; Shipley, W. U.; Urie, M. M.;McManus, P. L.; Efird, J.-T.; Willet, C. G.; Goitein, M. Laterectal bleeding following combined x-ray and proton highdose irradiation for patients with stage T3–T4 prostate carci-noma. Int. J. Radiat. Oncol. Biol. Phys.26:551–557; 1993.

5. De Neve, W., Van den Heuvel, F.; Coghe, M.; Verellen, D.;De Beukeleer, M.; Roelstraete, A.; Roover, P.; Thon, L.;Storme, G. Interactive use of on-line portal imaging in pelvic

Fig. 11. The probability distribution of a rectal volume receivingTD 5/5 of 60 Gy, obtained from setup deviations and the corre-sponding DVHs.


irradiation. Int. J. Radiat. Oncol. Biol. Phys. 25:517–524;1993.

6. Dearnaley, D. P.; Nahum, A. E.; Lee, M.; Tait Wynne, C.;Horwich, A. Radiotherapy of prostate cancer. Reducing thetreated volume. Vonformal therapy, hormone cytoreductionand protons (Abstract). Br. J. Cancer 70(Suppl XXII):16;1994.

7. Dearnaley, D. P.; Shearer, R. J.; Ellingham, L.; Gadd, J.;Horwich, A. Rationale and initial results of adjuvant hormonetherapy and irradiation for prostate cancer. In: Motta, M.;Serio, M. ed. Sex hormones and antihormones in endocrinedependent pathology: Basic and clinical aspects. Amsterdam:Elsevier Science B.V.; 1994.

8. Forman, J. D.; Zinreich, E.; Ding-Jen, L.; Wharam, M. D.;Baugardner, R. A.; Order, S. E. Improving the therapeuticratio of external beam irradiation for carcinoma of the pros-tate. Int. J. Radiat. Oncol. Biol. Phys. 11:2073–2080; 1985.

9. Gilhuijs, K. G. A.; Drukker, K.; Tonw, A., van de Ven,P. J. H.; and van Herl M. Interactive three-dimensional in-spection of patient setup in radiation therapeutic using digitalportal images and computed tomography data. Int. J. Radiat.Oncol. Biol. Phys. 34(4):873–885; 1996.

10. Gilhuijs, K. G. A.; van Herk, M. Automatic on-line inspectionof patient setup in radiation therapy using digital portal im-ages. Med. Phys. 20:667–677; 1993.

11. Gilhuijs, K. G. A.; van Herk, M.; Bijhold J. Automatic on-linepatient setup analysis in portal images. Med. Phys. 20:667–677 (1993).

12. Hanley, J.; Mageras, G. S.; Sun, J.; Kutcher, J. G. The effectsof out-of-plane rotation on two dimensional portal imageregistration in conformal radiotherapy of the prostate. Int. J.Radiat. Oncol. Biol. Phys. 33:1331–1343; 1995.

13. Herman, M. G.; Abrams, R. A.; Mayer, R. R. Clinical use ofon-line portal imaging for daily patient treatment verification.Int. J. Radiat. Oncol. Biol. Phys. 28:1017–1023; 1994.

14. Hunt, M. A.; Kutcher, G. J.; Burman, C.; Fass, D.; Harrison,L.; Leibel, S.; Fuks, Z. The effect of setup uncertainties on thetreatment of nasopharynx cancer. Int. J. Radiat. Oncol. Biol.Phys. 27:437–447; 1993.

15. Kutcher, G. J.; Mageras, G.; Georgiades, A. Setup errors andtarget margins: An uncertainty analysis. In: Proceedings of the3rd international workshop on electronic portal imaging. SanFrancisco: Manitoba Cancer Treatment & Research Founda-tion; 1994.

16. Lebesque, J. V.; Kues, R. The simultaneous boost technique:The concept of relative normalised total dose. Radiat. Oncol.22:45–55; 1991.

17. Lebesque, J. V.; Bruce, M.; Guus Kroes, A. P.; Touw, A.;Shouman T.; van Herk, M. Variation in volumes, dose–vol-ume histograms, and estimated normal tissue complicationprobabilities of rectum and bladder during conformal radio-therapy of T3 Prostate cancer. Int. J. Radiat. Oncol. Biol.Phys. 33:1109–1119; 1995.

18. Leibel, S. A.; Heimann, R.; Kutcher, G. J.; Zelefsky, M. J.;Burman, C. M.; Mellan, E.; Orazem, J. P., Mohan, R.;LoSasso, T. J. Three-dimensional conformal radiation therapyin locally advanced carcinoma of the prostate: Preliminaryresults of a Phase I dose–escalation study. Int. J. Radiat.Oncol. Biol. Phys. 28:55–65; 1994.

19. Mageras, G. S.; Kutcher, G. J.; Leibel, S. A.; Zelefsky, M. J.;Melian, E.; Mohan, R.; Fuks, Z. A method of incorporatingorgan motion uncertainties into three-dimensional conformaltreatment plans. Int. J. Radiat. Oncol. Biol. Phys. 35:333–342;1996.

20. Mubata, C. D.; Bidmead, A. M.; Ellingham, L. M.; Dearnaley,D. P. Incorporating random patient setup errors in 3D treat-ment planning. In: Proceedings of the 4th international con-

ference on electron portal imaging. Amsterdam: Free Univer-sity; 1996.

21. Niemierko, A.; Goitein, M. Implementation of a model forestimating tumor control probability for an inhomogeneouslyirradiated tumor. Radiat. Oncol. 29:140–147; 1993.

22. Padhani, A. Royal Marsden Hospital, Surrey, UK. Privatecommunication; 1996.

23. Rabinowitz, I.; Broomberg, J.; Goitein, M.; McCarthy, K.;Leong, J. Accuracy of radiation field alignment in clinicalpractice. Int. J. Radiat. Oncol. Biol. Phys. 11:1857–1867;1985.

24. Sandler, H.; McLaughlin, P. W.; Ten Haken, R.; Addison, H.;Forman, J.; Litcher, A. 3-D conformal radiotherapy for treat-ment of prostate cancer: Low risk of chronic rectal morbidityobserved in a large series of patients (Abstract). Int. J. Radiat.Oncol. Biol. Phys. 27(Suppl 1):135; 1993.

25. Sandler, H.; Perez-Tomayo, C.; Ten Haken, R.; Litcher, A.Dose escalation for Stage C (T3) prostate cancer: Minimalrectal toxicity observed using conformal therapy. Radiat. On-col. 23:53–54; 1992.

26. Shearer, R. J.; Davies, J. H.; Gelister, J. S. K.; Dearnaley,D. P. Hormonal cytoreduction and radiotherapy for carcinomaof the prostate. Br. J. Urol. 6:521–524; 1992.

27. Song, P. Y.; Washington, M.; Vaida, F.; Hamilton, R.; Spel-bring, D.; Wyman, B; Harrison, J; Chen, G. T. Y.; Vijayaku-mar, S. A comparison of four patient immobilisation devicesin the treatment of prostate cancer patients with three dimen-sional conformal radiotherapy. Int. J. Radiat. Oncol. Biol.Phys. 34:213–219; 1996.

28. Stavrev, P. V.; Stavreva, N. A.; Round, W. H. A new methodfor optimum dose distribution determination taking tumourmobility into account. Phys. Med. Biol. 41:1670–1689; 1996.

29. Ten Haken, R. K.; Forman, J. D.; Himburger, D. K.; Gerhard-son, A.; McShan, M. S.; Perez-Tamayo, C.; Schoeppel, S. L.;Lichter, A. S. Treatment planning issues related to prostatemovements in response to differential filling of the rectum andbladder. Int. J. Radiat. Oncol. Biol. Phys. 20:1317–1324;1991.

30. Ten Haken, R. K., Lam, K., Balter, J.; Schewe, J.; Hornick,D.; Hetrick, J.; Kota, S.; McShan, D. L. Daily patient reposi-tioning using room-based digital diagnostic x-ray imaging anda computerised treatment table with six degrees of freedom.In: Proceedings of the 4th International conference on electronportal imaging. Amsterdam: Free University; 1996.

31. Thompson, V.; Bimead, A. M.; Mubata, C. D. Pictorial re-view: Comparison of portal images and megavoltage verifi-cation films for conformal pelvic irradiation. Br. J. Radiol.69:1191-1195.

32. van Herk, M.; Bel, A.; Gilhuijs, K. G. A.; Vijlbrief, R. E. Acomprehensive system for the analysis of portal images. Ra-diat. Oncol. 29:221–229; 1993.

33. van Herk, M.; Bruce, A.; Guus Kroes, A. P.; Shouman, T.;Touw, A.; Lebesque, J. V. Quantification of organ motionduring conformal radiotherapy of the prostate by three dimen-sional image registration. Int. J. Radiat. Oncol. Biol. Phys.33:1311–1320; 1995.

34. Vijlbrief, R.; de Boer, R.; Gilhuijs, K.; Lebesque, J. V.; vanHerk, M. Implementation of electronic portal imaging in dailyclinical practice. In: Proceedings of the 4th Internationalworkshop on electronic portal imaging. Amsterdam: Free Uni-versity; 1996.

35. Zelefsky, M. J.; Leibel, S. A.; Burman, C. M., Kutcher, G. J.;Harrison, A.; Happersett, L.; Fuks, Z. Neoadjuvant hormonaltherapy improves the therapeutic ratio in patients with bulkyprostate treated with three dimensional conformal radiationtherapy. Int. J. Radiat. Oncol. Biol. Phys. 29:755–761; 1994.


Documents

Portal Imaging Protocol for Radical Dose-Escalated Radiotherapy Treatment of Prostate Cancer