Accuracy in target localization in stereotactic radiosurgery

Medical Dosimetry. Copyright

0 1997 American

Vol. 22, No. 1, pp. 53-58.

1997

Association of Medical Dcsimetrists Printed in the USA. All rights reserved

@x8-3947/97

$17.00 + .oo

PII SO958-3947(96)00151-3

ELSEVIER

ACCURACY DONG-RAK

IN TARGET LOCALIZATION RADIOSURGERY

CHOI, PH .D., ’ YONG

CHAN AHN,

IN STEREOTACTIC

M.D.,

’ DAE YONG

KIM,

M.D., ’

IL LEE, M.D.2 Departmentsof ‘RadiationOncology and 2Neurosurgery,SamsungMedical Center, Kangnam-Ku,Seoul, 135-230,Korea

SEIJNG JAE HUH,

M.D.’ and JUNG

Abstract-The accuracy in target localization of CT, MRI, and digital angiography and the isocentric deviation of linear accelerator were investigated for stereotactic radiosurgery. Twenty-five slice images using CT and MRI were obtained out of geometrical phantom which was designedto produce exact 3dimensional coordinates of several points within a 0.1 mm error range. These diagnostic imageswere transferred to a 3-D treatment planning systemthrough ethernet. Measured 3-D coordinatesof theseimages from the planning systemwere compared to known valuesby geometrical phantom. Anterior-posterior and lateral films were taken by digital angiography for measurementof spatial accuracy. The accuracy of gantry isocenter aligned by laser localizer was also measured.Lead ball was located at the isocenter of linear accelerator and x-ray films were taken by 4-MV photon beam with gantry anglesof 0, 90, 180 and 270 degrees.Overall procedure from CT scanningto treatment wascarried out using the geometrical phantom. The target accuracy was verified by high energy x-ray portal film. The accuracy of diagnostic machines were within 2.1 mm except MR-axial images.In caseof linear accelerator, the deviation of isocenter was within 0.7 mm. Finally, the total isocentrlc deviation of overall procedure was 1.3 2 0.5 mm (using CT localization). 0 1997American Association of Medical Dosimetrists Key Words: Stereotactic radiosurgery,

Accuracy in localization,

INTRODUCTION

4.MV photon beam, Geometrical phantom.

and so on.‘-” In general, limits of accuracy reflect technical limitations of the frames, treatment and diagnostic units. This study is to report the accuracy in target localization of our CT, MRI, and angiography and the isocentric deviation of linear acceIerator for stereotactic radiosurgery.

Stereotactic radiosurgery was first developed by Lek-

sell in the late 1940s to destroy dysfunctional loci in the brain using orthovoltage X-rays.’ Procedure of stereotactic radiosurgery consists of two-steps; defining the shape and location of the lesion in a stereotactic coordinate system using a special frame with CT, MRI or angiography; and delivering the planned radiation treatment. This treatment technique produces a concentrated focal irradiation to the lesion with steep dose gradients around the lesion. The rapid dose fall-off from the edge of the treatment volume can dramatically spare normal brain from radiation damage.2-5 As for conventional external beam therapy for the CNS, treatment-to-treatment variation of target localization is 3 mm, and the average discrepancy between the simulation films and port films is 5 mm with the mean deviation exceeding 7 mm.6 The advantage of stereotactic localization and treatment is improved,accuracy. There are many factors of uncertainty in stereotactic radiosurgery procedure including stereotactic frame, isocentric alignment, diagnostic image resolution, tissue motion,

MATERIALS

AND METHODS

Our SRS hardware uses a 4 MV photon beam from linear accelerator (CL6OOC, Varian Co., USA). A cylindrical collimation cone, adapted with cone sizes from 4 to 42 mm at the isocenter is used. Rectilinear Phantom Pointer (RLPP) and Laser Target Localizer Frame (LTLF) with precision translational motion are used to simulate the target coordinates (Fig. 1). The RLPP can be attached to our treatment couch. The center of the pointer coincides with the coordinates of the Brown-Roberts-Wells (BRW) headframe. Image data acquisition using CT, MRI, and angiographs Intermediate head ring and localizer for CT/MRI (GE HighSpeed Advantage/GE Signa Advantage 1.5T) were attached to the geometrical phantom of several objects of well-known coordinates and scanned using CT/MRI. Slice separation of 3 mm was chosen for this study. In case of MRI, slice view of sagittal and coronal sections were added to the axial section.

Presented at the 11 th annual meeting of the Korean Association of Physicists in Medicine (KAPM), Seoul, Korea, September 2728, 1995. Reprintrequests to: Dong-Rak Choi, Ph.D. Department of Radiation Oncology, Samsung Medical Center, 50 Ilwon-Dong, Kangnam-Ku, Seoul 13.5-230, Korea. E-mail address: [email protected] 53

Medical Dosimetry

Volume 22, Number 1, 1997

Fig. 1. Rectilinear Phantom Pointer (RLPP) and Laser Target Localizer Frame (LTLF) . They are used to verify 1) The correct setup of the BRW treatment coordinates with respect to isocenter ; 2) The alignment of the radiation and geometric isocenter with respect to the lasers; 3) The correct setup of the patient at the BRW treatment coordinates.

SW-AL angiographic localizer was used to estimate localization accuracy of our digital angiography (GE Advantx LCA). AP and LAT films were required to determine the 3-dimensional coordinates of geometrical objects. Determination phantom

for object coordinates

in geometrical

arc. The deviation of gantry rotation axis is the only factor to affect the accuracy in localization. The isocentric accuracy of the linear accelerator was tested for four gantry angles: 0, 90, 180, and 270 degrees. An X-omat film was located below lead ball which lay on the isocenter, and images were taken by irradiation. The developed images of lead ball and exposures were used to estimate the accuracy of gantry isocenter (Fig. 2).

There are four known coordinates in geometrical phantom. The geometrical images scanned by CT/MRI were tranferred to the treatment planning computer using ethernet lines. Three dimensional reconstructions of geometrical phantom were made by digitizing the slice images. In digital angiography, the coordinates of two orthogonal films were transferred to the treatment planning computer using a film scanner. The three dimensional value for each coordinate was determined by the treatment planning computer. isocentric

accuracy

of linear accelerator

Our SRS system is independent of couch rotation axis because the couch isocenter can be adjusted for each

Fig. 2. Verification films illustrating gantry alignment. Cone diameter is 12 mm and gantry angles are 0, 90, and 270 degrees.

Accuracy in target localization in stereotactic radiosurgery 0 D.-R.

CHOI

55

e6 nl.

tween phantom base measurementsand computer calculations, respectively.6 The components of the localization errors of CT are represented in Figs. 4 (a) and (b) . As shown in Fig. 4, the standard deviations of A AP, ALat, and Avert were 0.6 mm, 0.5 mm, and 1.0 mm, respectively. The use of CT scans with 512 x 512 matrix and 3 mm slice thickness resulted in a mean localization error of 1.2 f 0.5 mm for quadraplicate measurements of the four targets. This value was comparable to results of other studieslo (0.9 1 + 0.3 mm for 2-mm slice and 1.58 +- 0.5 mm for 4-mm slice). The accuracy of CT localization is directly related to 3-dimensional

3

T VERT

Fig. 3. Verification of overall accuracy using geometrical phantom.Verification films were exposedat gantry angles at 0 and 90 degrees.

-3 l a Overall accuracy in SRS procedure Total deviation of the isocenter was estimated by the previous overall procedure stepsusing geometrical phantom. The images of geometrical phantom were taken by CT/MRI or angiography. During theseprocedures, BRW head rings and CT/MRI or angiography localizers were attached to the phantom. These images were transferred to the treatment planning computer to get the 3-dimensional coordinates of each object. The determined coordinates were set to the isocenter of the treatment machine using a laser target localizer (Fig. 3). The portal images taken for gantry angles of 0 and 90 degree were analyzed to estimate accuracy in localization.

RESULTS

m t

AND DISCUSSION

CT localization In general, the localization error (Ar) as follows: Ar = d(AAP)’

3 -j’- VERT

+ (ALat)

where AAP, ALat, and Avert

b

is defined

+ (Avert)’

are the differences be-

Fig. 4. The componentsof spatialuncertainty of CT (unit: mm). (a) The AP-directional and VERT-directional differencesbetweenmeasurements and known coordinatevalues of geometrical phantom. (b) The LAT-directional and VERT-directional differencesbetween measurements and known coordinatevalues of geometricalphantom.

56

Medical Dosimetry

3.0

I

Volume 22, Number 1, 1997

graphic unit, the image warping was observed at the periphery of the imaging field. This finding suggests that the target should be located at the center of the imaging field.

VERT 0

Accuracy in gantry isocenter localization The components of the isocentric deviation of the treatment machine, including laser localization error are shown in Fig. 7. By analyzing 18 treatment portals of the targets, we found that the standard deviations of A AP, ALat, and Avert were 0.2 mm, 0.1 mm, and 0.1 mm, respectively. The error of vertical coordinates

L-l

MR-axial o MReoronal 0 MR-sa ittal n

I -6

3.0 - VERT 0 -0 0 1L ,=,2 . -3 . . 3.0!-

3 -

4 LAT 3

0

VERT

l --

. l

b Fig. 5. The componentsof spatialuncertainty of MR-axial, coronaland sagittalimages(unit: mm). (a) The AP-directional and VERT-directional differencesbetweenmeasurementsandknown coordinatevaluesof geometricalphantom. (b) The LAT-directional andVERT-directional differences betweenmeasurements andknown coordinatevaluesof geometrical phantom.

-3

3

-3 ~ image resolution. Therefore, in our SRS practice, we use the CT scan of 5 12 X 5 12 matrix and 1 mm slice

thickness for the target regions and 3 or 5 mm slice thickness for others.

a

3

T

VERT

MRI localization The components of localization errors of MRI are represented in Figs. 5(a) and (b). As shown in Fig. 5, the standard deviations of AAP, ALat, and Avert were within 1.4 mm. The use of MRI scans with 256 X 256 matrix and 3 mm slice thickness resulted in a mean localization error of 1.7 + 0.4 mm for MRcoronal and 2.1 + 0.7 mm for MR-sagittal. The accuracy of localization was very poor (mean value: 4.7 + 0.9 mm) for MR-axial; However, these errors can be controlled because the distortion is always shifted to the posterior. Angiographic localization The components of the localization errors of angiography are represented in Figs. 6(a) and (b). As shown in Fig. 6, the standard deviations of A AP, ALat, and Avert were 0.4 mm, 0.9 mm, and 0.3 mm, respectively. The use of digital angiography resulted in a mean localization error of 0.9 2 0.4 mm for duplicate measurements of the four targets. In our digital angio-

b Fig. 6. The componentsof spatialuncertainty of digital angiography (unit: mm). (a) The AP-directional and VERTdirectional differencesbetween measurements and known coordinatevalues of geometricalphantom. (b) The LATdirectional and VERT-directional differences between measurements and known coordinatevaluesof geometrical phantom.

Accuracy in target localization in stereotactic radiosurgery 0 D.-R. CHOI ef al. Table 1. Average localization

‘1VERT [:I

errors (Ar) in SRS Error (mm)

Procedure CT localization

1.2 4.1 1.7 2.1 0.9 0.5

MR-axial

coronal sagittal Angiography

Isocentric alignment

CONCLUSION

-1 1. a

1-

VERT Odeg El0 180deg l

. . . t

-1

0. . 4c .

I

II 0 000 0 0 -- 0 0

I LAT 1

The accuracy in localization of diagnostic and therapeutic machineswas estimated for stereotactic radiosurgery. A geometrical phantom of known coordinates within 0.1 mm was used to determine the localization ability of CT, MRI, and angiography in addition to 3-D treatment planning systemsfor SRS. The accuracy of gantry isocenter aligned by laser localizer was also estimated. Overall procedure from CT scanning

VERT

3T

-1 b Fig. 7. The components of isocentric deviation of the treatment machine (unit: mm). (a) The AP-directional and VERT-directional errors using gantry angle of 90 and 270 degrees. (b) The LAT-directional and VERT-directional errors using gantry angle of 0 and 180 degrees.

is due to the gantry

-31

for the gantry weight. At 0 degree of gantry, the mean isocentric deviation is 0.6 mm in the positive vertical direction. As the gantry angle increases,the isocentric deviation is shifted to the negative vertical direction (up to -0.5 mm). The spatial error in the treatment field center attributable solely to the setup was 0.5 -t 0.2 mm. sag to compensate

Overall accuracy in whole treatment procedure A summary of the mean localization error (Ar) in each SRS procedure using geometrical phantom is given in Table 1. The mean accuracy in localization of CT and digital angiographic units was 1.2 mm and 0.9 mm. respectively. These values were comparable to those of other studies.6-‘” The components of the total isocentric deviation of the overall procedure are shown in Fig. 8. In this case, 16 targets were localized by using CT (3-mm slice) with a localization error (Ar) of 1.3 4 0.5 mm. This result agrees with the accuracy estimated by Lutz et al.? and Yeung et al.”

a

38

. t

3

II

pisg

.

. --

. II

VERT a

n

mm.

. n

I

I I

i LAT 3

. I-3 1 b

Fig. 8. The componentsof cumulative error in treatment delivery using CT (unit: mm). (a) The AP-directional VERT-dirctional deviation. (b) The LAT-directional VERT-directional deviation.

and and

Medical

58

Dosimetry

to treatment was carried out using geometrical phantom. The mean localization errors of diagnostic machines were 0.9 to 2.1 mm except in MR-axial images. In case of isocentric alignment, the mean spatial error was 0.5 + 0.2 mm. The accuracy of our SRS system is mainly dependent on CT localization, because CT images are always the primary standard for localization and the isocentric deviations of treatment machines are relatively small. The information on total isocentric deviation of the overall procedure was very useful in determining a target margin. Using CT, the localization error of overall procedure was 1.3 t 0.5 mm. These results agree with the accuracy estimated by others in previous studies.6Z7

REFERENCES 1. Leksell, L.T. The stereotactic method and radiosurgery of the brain. Actu Chir. Stand. 102:316-319; 1951. 2. Arcovito, G.; Piermattei, A.; D’Abramo, G.; Bassi, F.A. Dose

Volume

22, Number

1, 1997

measurements and calculations of small radiation fields for 9MV x-rays. Med. Phys. 12:779-784; 1985. 3. Lutz, W.; Winston, K.R.; Maleki, N. A system for stereotactic radiosurgery with a linear accelerator. ht. J. Radiat. Oncol. Biol. Phvs. 14:373-381; 1988. 4. Winston, K.R.; Lutz, W. Linear accelerator as a neurosurgical tool for stereotactic radiosurgery. Neurosurge~. 22:454-463; 1988. 5. Bjarngard, B.E.; Tsai, J.S.; Rice, R.K. Doses on central axis of narrow 6-MV x-ray beams. Med. Phys. 17:794-799; 1990. 6. AAPM Stereotactic Radiosurgery AAPM Report 54; 1995. 7. Hartmann, G.H.; Bauer-Kirpes, B.; Serago, C.F.: Lorentz, W.J. Precision and accuracy of stereotactic convergent beam irradiations from a linear accelerator. Znt. J. Radiat. Oncol. Biol. Phys. 28:481-492; 1993. 8. Rabinowitz, 1.; Broomberg, J.; Goitein, M.; Goitein, M.: McCar thy, K.; Leong, J. Accuracy of radiation field alignment in clinical practice. Znt. J. Radiut. Oncol. Biol. Phys. 11:18571867; 1985. 9. Serago, C.F.; Lewin, A.A.; Houdek, P.V.; Gonzales-Arias, S.; Hartmann, G.H. Stereotactic target point verification of an x-ray and CT localizer. Int. J. Radiut. Oncol. Biol. Phvs. 20:5 17-523; 1991. 10. Yeung, D.; Palta, J.; Fontanesi, J.; Kun, L. Systematic analysis of errors in target localization and treatment delivery in stereotactic radiosurgery (SRS). Int. J. Radiat. Oncol. Biol. Phys. 28: 493-498; 1994.