isocenter accuracy in frameless stereotactic radiotherapy using implanted fiducials

doi:10.1016/S0360-3016(03)00088-9

3D-CRT

ISOCENTER ACCURACY IN FRAMELESS STEREOTACTIC RADIOTHERAPYUSING IMPLANTED FIDUCIALS

KI-HWAN KIM, M.S.,* MOON-JUNE CHO, M.D.,*† JUN-SANG KIM, M.D.,*† JAE-SUNG KIM, M.D.,*†

CHANG-JOON SONG, M.D.,‡ SHI-HUN SONG, M.D.,§ SEON-HWAN KIM, M.D.,§ LEE MYERS, M.D.,� AND

YONG-EUN KIM, PH.D.¶

Departments of *Therapeutic Radiology;‡Diagnostic Radiology; and§Neurosurgery, College of Medicine;†Cancer ResearchInstitute, Chungnam National University, Taejon, South Korea;�Northwest Medical Physics Equipment, Lynnwood, WA;

¶Department of Physics, College of Natural Science, Chungbuk National University, Cheongju, South Korea

Purpose: The stereotactic radiotherapy (SRT) system verifies isocenter accuracy in patient space. In this study,we evaluate isocenter accuracy in frameless SRT using implanted cranial gold markers.Methods and Materials: We performed frameless SRT on 43 intracranial tumor patients between August 1997and December 2000. The treatment technique was determined by the tumor shape and volume, and by thelocation of critical organs. The coordinates of anterior-posterior and lateral port film were inputted to ISOLOCsoftware, which calculated (1) the couch moves translation distance required to bring the target point to theisocenter, and (2) the intermarker distance comparisons between the CT study and the treatment machine films.We evaluated the isocenter deviation based on the error between orthogonal film target coordinates and isocentercoordinates.Results: The mean treatment isocenter deviations (x, y, z) were �0.03, 0.14, and �0.04 mm, respectively. Thesystematic component isocenter standard deviations were 0.28, 0.31, and 0.35 mm (1 SD), respectively, and therandom component isocenter standard deviations were 0.53, 0.52, and 0.50 mm (1 SD), respectively.Conclusion: The isocenter accuracy in the frameless SRT–implanted fiducial system is highly reliable and iscomparable to that of other stereotactic radiosurgery systems. © 2003 Elsevier Inc.

Frameless, Stereotactic radiotherapy, Stereotactic radiosurgery, Fractionation, Isocenter, Accuracy.

INTRODUCTION

Stereotactic radiosurgery (SRS) is defined as treatment inwhich a large single dose is delivered using rigid immobi-lization with accuracy to within 1 mm in stereotactic space.SRS is indicated for both benign and malignant intracraniallesions. However, reported differential radiobiology be-tween dose–response curves of tumor control and late nor-mal-tissue effects favors dose fractionation (1). Late-re-sponding normal-tissue effects from high single-dose SRShave led to interest in the alternative of high-accuracyfractionated stereotactic radiotherapy (SRT). Invasive SRShead frames rigidly fix the geometry of intracranial lesioncoordinates. Noninvasive SRT systems employing masks(2–4), relocatable frames (5–8), and implanted fiducialmarkers (9) have been developed to achieve a high degreeof accuracy.

Isocenter accuracy for SRT is essentially equivalent tothat of SRS, where accuracy is generally obtained to within

1 mm. However, SRT systems allow a greater probability ofhead motion than with a fixed frame. Accordingly, a methodof monitoring and verifying stereotactic repositioning accu-racy is needed. In this study, we report isocenter accuracy ina frameless SRT system (pReference System, NMPE, Lynn-wood, WA) using implanted cranial gold markers and ther-moplastic mask technology.

METHODS AND MATERIALS

Between August 1997 and December 2000, 43 intracra-nial tumor patients were treated with frameless SRT at theChungnam National University Hospital.

Imaging studyTo define the stereotaxis, 3 gold markers 2.0 mm in

diameter were implanted in the cranium. The location ofmarkers is not critical, but 2 of 3 markers were generallyplaced on the left and right sides of the coronal suture, and

Reprint requests to: Moon-June Cho, M.D., Department ofTherapeutic Radiology, Chungnam National University Hospital,301-721, 640 Daesa-Dong, Jung-gu, Taejon, South Korea. Tel:82-42-220-7861; Fax: 82-42-220-7899; E-mail: [email protected]

Presented at the 3rd S. Takahashi Memorial International Work-

shop on 3-Dimensional Conformal Radiotherapy, December 8–10,2001, Nagoya, Japan.

Received Feb 22, 2002, and in revised form Aug 7, 2002.Accepted for publication Aug 23, 2002.

Int. J. Radiation Oncology Biol. Phys., Vol. 56, No. 1, pp. 266–273, 2003Copyright © 2003 Elsevier Inc.

Printed in the USA. All rights reserved0360-3016/03/$–see front matter

266

the third in the midanterior hairline region by the neurosur-geon according to surgical protocol for implantation (Fig.1). After 3 or 4 days, the patient was immobilized in thestandard head holder provided with the CT scanner. Therewas no requirement to fix a particular head position, becausethe target was localized with respect to the gold markers.After the gold markers in the anterior-posterior and lateralscout views were verified on the CT console monitor, aseries of image scans of patients was started to define themarker coordinates on the skull. A slice thickness of 1 mmwas used to obtain high spatial resolution.

The i.v. contrast media (Ultravist) was injected at a doseof 200 cc to identify the tumor region. Target volume sliceswere acquired at a slice thickness of 1 mm and a spacing of3 mm. The upper and lower general anatomy series wereacquired at a thickness of 1 mm and a spacing of 10 mm.

The CT study was reviewed to ensure that patients wereimmobilized during scanning and that the coordinates ofeach of the markers were identified. Figure 2 shows a pageof images documenting the locations of the target andmarkers. The image matrix size is 256 � 256, and the fieldof view is 25 � 25 cm (2).

Simulation and verification of patient positioningFrameless SRT entails establishing the relationship be-

tween marker coordinates in the treatment machine and CTcoordinate systems; accordingly, the treatment position isdefined at the simulation session on the radiotherapy ma-chine, a Clinac 2100C/D (Varian, USA) (Fig. 3). Afterimmobilization is achieved with a thermoplastic mask andback supporter, a cassette holder is attached to the linactable for anterior-posterior and lateral beam films (ECL,Kodak, USA). Patient cooperation is of the utmost impor-tance in this procedure, and we have found it advantageousto make the patient as comfortable as possible to promoteimmobilization. A small-field SRT collimator is attachedinto the wedge slot of the machine. Index marks defining theaxis of the radiotherapy machine are built into the collima-tor assembly, and a “4” sign is built into the collimator

assembly to verify the magnification scale on the port film.When taking port films, irradiation field size is set to 30 �30 cm2 with 3 to 4 monitor units.

After the beam films are produced, the coordinates of theimplanted marker of the treatment machine are defined bydigitization. A micropositioner is attached to the cassetteholder and set to zero position on each of 3 axes. A pointeron a floor stand is then aligned with the micropositioner.After this, the couch translation movements required tobring the target point to the isocenter are dialed into themicropositioner, and the couch is carefully moved to bringthe pointer on the micropositioner back into alignment withthe pointer on the floor stand. The patient structure may berotated around all axes and placed in a coordinate systemfixed in the treatment room with the origin at the isocenter.However, the rotated pattern of the isocenter is close to thatof the translated component in the target region. Because thepatient’s maximum anatomy rotation errors do not exceed�5° in our case, we believe that the movement of therotational component can effectively be ignored. Two filmsare taken at known gantry angles (0° and 90°) specified inthe International Electrotechnical Commission (IEC) sys-tem, and the images of the 1 to 3 markers are located. Thex, y (H, V) coordinates on the film are determined for eachof the images; from these, the location of the target and thetranslation of the couch required to bring it to the isocenterare calculated, and the intermarker distance comparisonsbetween the coordinates in the CT study and the coordinatesin the treatment machine films are shown with respect to 3gold markers using ISOLOC software (NMPE, USA). Themethod depends on the principle that if the coordinates of 3reference points and the coordinates of a target point areknown in any initial coordinate system, and if the coordi-nates of 3 points are determined in another coordinatesystem at a subsequent time (e.g. by an isocentric linac),then the coordinates of the target may be correspondinglycalculated in this second system. After this, the 3 markerscan be used to transpose the coordinates of a fourth point(i.e., target point) into an isocenter. The criterion for ste-

Fig. 1. The protocol for implanting 3 gold markers on the skull.

267Isocenter accuracy in SRT using implanted fiducials ● K.-H. KIM et al.

reotactic alignment is a translational variation between tar-get and isocenter of less than 1 mm in each axis.

Isocenter setup error analysisThe error was investigated on 3 axes (x, y, and z) using

three categories (overall, systematic, and random) accordingto El-Gayed et al. (10). We defined the mediolateral direc-tion as the x axis, the anterior-posterior direction as the yaxis, and the craniocaudal direction as the z axis.

The positive direction is defined as the left anterior cra-nial direction in the room coordinate system with the supineposition.

For a given axis, the overall displacement was defined asthe mean of all setup displacements for all patients. Thedistribution of the overall displacements was determined bythe standard deviation.

For a given axis, the systematic component displace-ment, representing persistent positioning variation for anindividual patient, has been obtained from the meandisplacement along certain coordinates for that individualpatient. An estimate is obtained of the distribution of thesystematic component displacements in the setup for allpatients, determined by the standard deviation (1 SD) ofthe mean shifts of individual patients along specific co-ordinates. For a given axis, the random component dis-placements, representing day-to-day variation, were de-fined as those displacements determined by subtraction ofthe systematic component displacement from all setupdisplacements in the simulation session. The distributionof the random component displacements was determinedby calculating the standard deviation (1 SD) of the indi-vidual displacement variations.

Fig. 2. The coordinates of 3 implanted cranial markers: (a–c) Define stereotaxy for intracranial lesions; (d) Showcoordinates of the target in CT space.

268 I. J. Radiation Oncology ● Biology ● Physics Volume 56, Number 1, 2003

RESULTS

The average displacements of the isocenter in 3 directionswere �0.03 mm (SD 0.53 mm) in the mediolateral direction(x), 0.14 mm (SD 0.51 mm) in the anterior-posterior direc-tion (y), and �0.04 mm (SD 0.62 mm) in the craniocaudaldirection (z) (Table 1).The systematic component standarddeviations in the 3 directions were 0.28 mm on the x axis,0.31 mm on the y axis, and 0.35 mm on the z axis. Therandom component standard deviations in the 3 directionswere 0.53 mm on the x axis, 0.52 mm on the y axis, and 0.50mm on the z axis.

Figure 4 a shows a systematic component isocenter dis-placement histogram for the mediolateral axis, and Fig. 4bshows a random component isocenter displacement histo-gram for the mediolateral axis. The positive direction isdefined as left in the room coordinate system with thesupine position. The systematic component distribution inFig. 4a shows no asymmetry, whereas the random compo-nent distribution in Fig. 4b is generally shifted to the left.The random component distribution in Fig. 4b is muchbroader and closer to a normal distribution than the system-atic component distribution in Fig. 4a.

Figure 5 a shows a systematic component isocenter dis-

placement histogram for the anterior-posterior axis, and Fig.5b shows a random component isocenter displacement his-togram for the anterior-posterior axis. The positive directionis defined as anterior in the room coordinate system with thesupine position. The systematic component distribution inFig. 5a and the random component distribution in Fig. 5bare shifted in the anterior direction. The random componentdistribution tendency in the anterior-posterior direction isthe same as that in Fig. 5b. The random component distri-bution in Fig. 5b is much broader and closer to a normaldistribution than the systematic component distribution inFig. 5a.

Figure 6 a shows a systematic component isocenter dis-placement histogram for the craniocaudal axis, and Fig. 6bshows a random component isocenter displacement histo-gram for the craniocaudal axis. The positive direction isdefined as cranial direction in the room coordinate systemwith the supine position. The systematic component distri-bution shown in Fig. 6a is close to flat, whereas the randomcomponent distribution shown in Fig. 6b is generally shiftedwith symmetrical form in the cranial direction. The randomcomponent distribution tendency in the anterior-posteriordirection is the same as that in Fig. 4b or Fig. 5b.

Fig. 3. Patient setup in the simulation procedure that two orthogonal port films in a cassette holder slides attached linactable were taken to anterior-posterior and lateral beam films.

Table 1. Measured isocenter accuracy by the average displacement and the standard deviation ofthe distributions of overall, systematic, and random displacements

Overall displacement(mm)

Distribution of displacements (1 SD, mm)

Overall(n � 304)

Systematic(n � 43)

Random(n � 304)

Mediolateral (x) �0.03 0.53 0.28 0.53Anterior-superior (y) 0.14 0.51 0.31 0.52Craniocaudal (z) �0.04 0.62 0.35 0.50


The random component distributions do not differ fromnormal distributions in any of the three directions. Therandom component distribution magnitudes were in thesame range as those in the overall distributions, whereas therandom component distribution magnitudes were largerthan those of the systematic component distributions inthree directions.

DISCUSSION

The standard methodology in isocentric linac radiosur-gery has been to rigidly attach a localizing frame to thepatient’s head, to determine the location of the target withrespect to the frame, and to perform dose calculations todetermine the optimum field arrangement and machine set-tings to deliver the prescribed dose. Using the BRW stereo-tactic frame attached to the therapy machine for aligning thetarget with the machine isocenter, Lutz et al. (11) reporteda positional accuracy of 2.4 mm in any direction with 95%confidence. Patient head position adjustment in the frame isaffected without body position translation by correspondingtreatment couch movement. Lutz et al. (11) and Yeung et al.(22) reported improved accuracy when angiography is used,

indicating that the limitation in accuracy related to thelocalization process rather than to the fixation system ormechanical stability of the treatment machine.

Other stereotactic methods are noninvasive. Masks (2–4), relocatable frames (5–8), and ear canal systems (13)have been used for setup reproducibility. Less than 1–2mm of isocenter deviation in all directions has beenreported with the use of immobilization masks (2–4).Isocenter repositioning accuracy of less than 1 mm in alldirections has been reported with the use of a relocatableframe, an immobilization device that may be removedand reattached with adequate precision for the treatmentof small intracranial lesions (5–8). Denannes et al. (13)reported setup accuracy within 1 mm using an ear canalsystem. These systems are considered to require a carefulalignment of the fixation device to the isocenter, althoughLyman et al. (4) reported on the use of localizing radi-ography at the time of treatment to correct minor mis-alignment (0.5 mm).

A separate noninvasive stereotactic method is to im-plant cranial markers, which are visible in the targetimaging study. This method defines target location in thecoordinate system created by the markers and was

Fig. 4. (a) Systematic component isocenter displacement histogram for the mediolateral axis. (b) Random componentisocenter displacement histogram for the mediolateral axis. The positive direction is defined as left in the roomcoordinate system with the supine position.


described by Columbo et al. (14) and developed inde-pendently by Jones et al. (9, 15). For stereotaxis, thetarget point is defined with respect to the 3 cranialmarkers, the same 3 points defined with respect to thetreatment machine, and the couch movements required tobring the target point to the treatment machine isocenterare then calculated.

The advantages of this methodology are that the localizationprocedure and treatment can be well separated in time. Frac-tionation of the single large dose is thus readily accomplished,while the accuracy of all treatments is ensured. Furthermore,the treatment beam itself is used to localize the target withreference to the markers, as an alternative to the stereotactichead frame. Use of this method is very simple for not onlyintracranial lesions but extracranial lesions as well, because itis unnecessary to carefully align a stereotactic frame to local-ization equipment such as a CT scanner or to a treatmentmachine such as a linear accelerator.

We used the frameless SRT system developed by Jones etal. using implanted gold markers (9, 15). The authors re-ported on the use of a patient coordinate system in thebrachytherapy of brain tumors, and they have applied thesame principle to the problem of external beam stereotactic

irradiation. We used a thermoplastic mask for immobiliza-tion to prevent the movement of patients during simulationand treatment sessions, although the methodology does notrequire any particular type of immobilization. It is unnec-essary for patient head position to be consistent with respectto the treatment machine, because the target position isrecalculated for each treatment setup. The only criterion isthat patient movement be limited to less than 1 mm betweenthe time of the last pair of localization films and the end ofthe treatment, i.e., that the target center is maintained at theisocenter. We adjusted the couch position until the requiredcouch movement in any direction met this criterion. Veri-fication of immobilization and target position is essential inSRS and SRT dose delivery. Accurate target alignment tothe isocenter means dose distribution corresponding to treat-ment planning and prescribed isocenter dose.

For a single isocenter plan, Ebert et al. (16) reported thatsystematic error in target localization results in a translationof the dose distribution in the three-dimensional space rep-resented by the patient space. Thus, a thorough qualityassurance program is required to ensure minimal systematicerror, and sufficiently robust localization and stabilizationmethods are required to minimize random setup error.

Fig. 5. (a) Systematic component isocenter displacement histogram for the anterior-posterior axis. (b) Randomcomponent isocenter displacement histogram for the anterior-posterior axis. The positive direction is defined as anteriorin the room coordinate system with the supine position.


Other studies by Hess et al. (17) and Lebesque et al. (18)have suggested that most of the systematic errors that affectthe overall quality of the treatment can be detected on thefirst day of treatment, because field errors have been shownto be considerably more prominent between simulation andthe first check film than during the subsequent course oftreatment delivery.

Our data show the systematic component and randomcomponent displacement distribution accuracy (1 SD)within 1 mm for intracranial tumor setup displacements,although the random component displacement deviationwas greater than systematic component displacement devi-ation.

Our data show random component isocenter displace-ments to be more symmetrically distributed than aresystematic component displacements. Accordingly, webelieve random component inaccuracies may have a rel-atively greater effect in our frameless SRT dose delivery,in contrast to other reports of a relatively greater contri-bution from systematic component inaccuracies. In addi-tion, Fig. 4a is the histogram of systematic error thatshows relatively symmetric form with respect to medio-lateral (x) axis, and the distribution of random error inFig. 4b is generally shifted to the left direction. It is

indicated the error in x axis is stable in x axis. In Fig. 5a,the systematic error is shifted in the positive (anterior)direction with respect to the anterior-posterior (y) axis. InFig. 5b, the value of random error can be consideredrelatively more or less symmetric than the value of sys-tematic error on the y axis. The histogram in Fig. 6ashows that the distribution of random error is spreadthroughout a range; the degree of inaccuracy in thecraniocaudal (z) direction is considered to be higher thanthat on any other axis. In contrast, the histogram in Fig.6b shows the random error on the z axis to be more stablethan the systematic error on the z axis, although thedistribution of random error in Fig. 6b is generally shiftedto the cranial direction. The random component distribu-tions do not differ from normal distributions in any of thethree directions. On our SRT system, it seems the pat-terns of isocenter displacements shift toward the positivedirection (i.e., left, anterior, and cranial direction), andthe random component distribution magnitudes are largerthan those of the systematic component distributions inthree directions.

There are two aspects to this problem. Serago et al. (19)have estimated a 0.3-mm uncertainty in point localization,

Fig. 6. (a) Systematic component isocenter displacement histogram for the craniocaudal axis. (b) Random componentisocenter displacement histogram for the craniocaudal axis. The positive direction is defined as cranial in the roomcoordinate system with the supine position.


because of brain pulsation. We believe that cranial markerimplantation obviates pulsation error. The second localiza-tion aspect involves the potential for patient movementduring the treatment or (in frame-based stereotaxis) after thetarget is localized with respect to the frame.

We have attempted to maximize isocenter accuracy inframeless stereotactic radiotherapy. Our data indicate thatthe isocenter accuracy in our frameless SRT system usingimplanted fiducials for isocenter tracking is equivalent tothat of other SRS systems.

REFERENCES

1. Hall EJ, Brenner DJ. The radiobiology of radiosurgery ratio-nale for different treatment regimes for AVM’s and malignan-cies. Int J Radiat Oncol Biol Phys 1993;25:381–385.

2. Willer J, Flentje M, Bratengeier K. CT simulation in stereo-tactic brain radiotherapy—analysis of isocenter reproducibil-ity with mask fixation. Radiother Oncol 1997;45:83–88.

3. Karger CP, Jakel O, Debus J, et al. Three-dimensional accu-racy and interfractional reproducibility of patient fixation andpositioning using a stereotactic head mask system. Int J RadiatOncol Biol Phys 2001;49:1493–1504.

4. Lyman JT, Phillips MH, Frankel KA, et al. Stereotactic framefor neuroradiology and charged particle bragg peak radiosur-gery of intracranial disorders. Int J Radiat Oncol Biol Phys1989;16:1615–1621.

5. Gill SS, Thomas DGT, Warrington AP, et al. Relocatableframe for stereotactic external beam radiotherapy. Int J RadiatOncol Biol Phys 1991;20:599–603.

6. Kooy HM, Dunbar SF, Tarbell NJ, et al. Adaptation andverification of the relocatable Gill-Thomas-Cosman frame instereotactic radiotherapy. Int J Radiat Oncol Biol Phys 1994;30:685–691.

7. Laing RW, Thompson V, Warrington AP, et al. Feasibility ofpatient immobilization for conventional cranial irradiationwith a relocatable stereotactic frame. Br J Radiol 1993;66:1020–1024.

8. Rosenthal SA, Gall KP, Jackson M, et al. A precision cranialimmobilization system for conformal stereotactic fractionatedradiation therapy. Int J Radiat Oncol Biol Phys 1995;33:1239–1245.

9. Jones D, Christopherson DA, Washinton JT, et al. A framelessmethod for stereotactic radiotherapy. Br J Radiol 1993;66:1142–1150.

10. El-Gayed AAH, Bel A, Vijlbrief R, et al. Time trend of patient

setup deviations during pelvic irradiation using electronicportal imaging. Radiother Oncol 1993;26:162–171.

11. Lutz W, Winston KR, Maleki N. A system for stereotacticradiosurgery with a linear accelerator. Int J Radiat Oncol BiolPhys 1988;14:373–381.

12. Yeung D, Jatinder P, Fontanesi J, et al. Systematic analysis oferrors in target localization and treatment delivery in stereo-tactic radiosurgery. Int J Radiat Oncol Biol Phys 2001;51:545–554.

13. Denannes M, Duly NJ, Bonnet J, et al. Fractionated radiother-apy of small inoperable lesions of the brain using a non-invasive stereotactic frame. Int J Radiat Oncol Biol Phys1991;21:749–755.

14. Columbo F, Anggrilli F, Zanard O, et al. A universal methodto employ CT scanner spatial information in stereotactic sur-gery. App Neurophysiol 1982;45:352–364.

15. Jones D, Christopherson DA, Raisis JE, et al. The use of anexternal reference system in brain implants. Endocuri Hyper-ther Oncol 1989;5:221–226.

16. Ebert MA, Zavgorodni SF, Kendrick LA, et al. Multi-iso-center stereotactic radiotherapy: Implication for target dosedistributions of systematic and random localization errors. IntJ Radiat Oncol Biol Phys 2001;51:545–554.

17. Hess CF, Kortmann RD, Jany R, et al. Accuracy of fieldalignment in radiotherapy of head and neck cancer utilizingindividualized face mask immobilization: A retrospectiveanalysis of clinical practice. Radiother Oncol 1995;34:69–72.

18. Lebesque JV, Bel A, Bijhold J, et al. Detection of systematicpatient setup errors by portal film analysis. Radiother Oncol1992;23:201–210.

19. Serago CF, Lewin AA, Houdek PV, et al. Radiosurgery targetpoint alignment errors detected with portal film verification.Int J Radiat Oncol Biol Phys 1992;24:777–780.


isocenter accuracy in frameless stereotactic radiotherapy using implanted fiducials

Documents