- Email: [email protected]
New system for linear accelerator radiosurgery with a gantry-mounted video camera
Int. J. Radiation
Biol.
Phya., Vol. 40. No. 3. pp. 739-746. IYYX Copyright 0 199X Elsrvirr Science Inc. Printed in the USA. All right\ reserved 0360-3016/YX $I’).00 + .OO
PI1 SO360-3016(97)00844-4
ELSEVIER
l
Oncology
Physics Contribution NEW SYSTEM
FOR LINEAR ACCELERATOR GANTRY-MOUNTED VIDEO
ETSUO KUNIEDA, M.D.,* MASAYUKI KITAMURA, M.D.,* TAKAYUKI OHIRA, M.D.,’ KOUICHI OGAWA, PH.D.,*
KAYOKO NAKAMURA
PH.D.* AND ATSUSHI
RADIOSURGERY CAMERA OSAMU
A
M.D.,* M.D.,*
KAWAGUCHI,
YUTAKA ANDO, KUBO,
WITH
M.D.*
*Departmentsof Radiology,and ‘Neurosurgery,Schoolof Medicine, Keio University, Tokyo, Japan;and“Departmentof Electrical Engineering,Collegeof Engineering,HoseiUniversity. Tokyo, Japan Purpose: We developed a positioning method that does not depend on the positioning mechanism originally annexed to the linac and investigated the positioning errors of the system. Methods and Materials: A small video camera was placed at a location optically identical to the linac x-ray source. A target pointer comprising a convex lens and bull’s eye was attached to the arc of the Leksell stereotactic system so that the lens would form a virtual image of the bull’s eye (virtual target) at the position of the center of the arc. The linac gantry and target pointer were placed at the side and top to adjust the arc center to the isocenter by referring the virtual target. Coincidence of the target and the isocenter could be confirmed in any combination of the couch and gantry rotation. In order to evaluate the accuracy of the positioning, a tungsten ball was attached to the stereotactic frame as a simulated target, which was repeatedly localized and repositioned to estimate the magnitude of the error. The center of the circular field defined by the collimator was marked on the film. Results: The differences between the marked centers of the circular field and the centers of the shadow of the simulated target were less than 0.3 mm. 0 1998 Elsevier Science Inc. Stereotactic
radiosurgery,
Linear
accelerator,
Video camera, Beam&eye
INTRODUCTION
monitoring.
on laser positioners originally annexed to the machine for conventional radiotherapy, and we report our examination of the accuracy of this positioning method.
Although the gamma unit produced by Leksell (1, 2) has been applied for more than 2 decades and has demonstrated the clinical utility of radiosurgery (3-5) there has been increasing interest in carrying out stereotactic radiosurgery using x-ray beams produced by linear accelerators, or linacs (6-10). In addition to the quantitative advantages of linac-based radiosurgery over gamma units, such as low initial introductory cost and the absence of any need for source replacement, the linac’s increased source-collimator distance (8) reduces limitation in the size of the radiation field. On the other hand, most of the medical linacs currently in use were not originally designedto be applied to radiosurgery: the accuracy required for this purpose exceeds the level that the vendors of these therapy machineshave assumedtheir products to have. We describe in this paper a new radiosurgery system that employs a positioning method that does not depend
The linear accelerator and auxiliary devices A 6MV linear accelerator (MLlSMV: Mitsubishi Electries Corp., Tokyo, Japan) was used to produce the x-ray beam for stereotactic radiosurgery. We utilized secondary collimators with 5 mm to 30 mm-diameter openings projected to the isocenter of the machine. The height of the collimator is 110 mm and its isocenter-side end is positioned 320 mm away from the isocenter of the machine. During treatment, the base frame securing the patient’s head is fitted to the treatment couch by an adaptor equipped with three orthogonal drives and verniers.
Presented at First Korea-Japan Joint Meeting on Medical Physits. Seoul,Korea. September19. 1996. Reprint requeststo: EtsuoKunieda,M.D.. Departmentof Ra-
diology. Keio University, 35 Shinanomachi, Shinjuku.Tokyo 160 Japan. Acceptedfor publication25 September1997.
METHODS
739
I. J. RadiationOncology l Biology l Physics
140 X-ray
'\
source
Volume 40, Number 3, 1998
CCD video camera and the light source
Stereotactic
Target Stereotactic ar -
-..-*.. .._ -.._--. ... .. ...I----v
/-
arc
’
pointer
Convex lens Bull's eye
“Virtual target” arc center
at the c
Fig. 1. A schematic illustration of the positioning method. A small CCD video camera (beam’s-eye monitor) is mounted at a point optically identical to the linac x-ray source. The center of the arc is adjusted to the gantry isocenter under the guidance of the “virtual target” produced by a convex lens and a bullseye (target pointer) mounted on the stereotactic XC.
Beam ‘s-eye monitor In the gantry of a linac, a small CCD (charge-coupled device) video camera is mounted at the position where the source of the light field is usually placed. The light source and the video camera are alternately guided into the exact position when each of them is needed. The video camera is located at the position optically identical to the linac x-ray source, enabling real-time observation of the margin and the center of the irradiation field. Because the video camera displays the irradiation field as if viewed from the eye point of the x-ray beam source, we refer to it as a beam’s-eye monitor (Fig. 1). The video signal from the camera is digitized by means of the video-capture function of a personal computer (Macintosh 7100AV: Apple Computer Inc., Cupertino, CA). The matrix of the image is 640 X 480, with a pixel size of 0.14 X 0.14 mm. The images of the beam’s-eye monitor can be seen as motion images or static images at any moment. Arc system and target pointer Instead of depending on the beam’s alignment, we adopted a positioning method that is independent of the
room laser beams. The schematic shown in Fig. 1 gives a simple overview of our method. A positioning device, a target pointer, is attached to the arc of the stereotactic system. The arc is a part of the Leksell stereotactic system (Leksell Stereotactic system Type G, Elekta Instruments, AB, Stockholm, Sweden) and is used primarily to hold a needle for biopsy or aspiration of intracranial fluid. The target pointer slides along the arc with its axis kept perpendicular to the arc. A convex lens and bullseye are mounted on the target pointer in the same optical alignment along its long axis. Because the head of the patient is kept in the base frame, it is impossible to place into the machine isocenter an object that would indicate the arc center. Nevertheless, the lens makes a virtual image of the bullseye, a virtual target, at the position of the arc center, even though the actual position of the bullseye is far way from the center (Fig. 2). The target pointer pivots on the center of the arc, and we can turn the target pointer toward any direction from which the accelerator’s beam might come (Fig. 3). The stereotactic arc is mounted on the base frame so that the center of the arc can be matched with the intended target
Gantry-mountedvideo camera0
CCD camera L,
( f= f,)
l/a-l/b=l/ft
E. KUNIEDA
741
et al.
gles in a similar way to the definition of the mechanical isocenter. Because the largest displacements of the gantry are at the top and bottom positions of the gantry rotation (1 l), only the adjustment parallel to the gantry rotation plane is performed at the gantry’s top position. If collimators of less than 10 mm are used, these positioning procedures are repeatedly carried out in each couch rotation for an arc irradiation.
l/D+l/d=l/fc MEASUREMENTS
L+I=SAD
I I I
---
+L-_--d
Fig. 2. Optical geometry of the CCD camera and the target pointer when the arc center coincides with the isocenter of the gantry rotation. L, and L,., the lenses of the target pointer and the CCD camera, have focal distances off, and f,, respectively. The bullseye (B) and its virtual image (B’) are indicated in the figure. SAD = Source axis distance. point by referring to the Cartesian coordinates obtained by the localization procedure. The target point is roughly positioned at the isocenter of the linac using the conventional laser beam positioner in advance to obtain more precise positioning. Next, the linac gantry and the target pointer are positioned at 90” or 270” (side position), so that the image of the bullseye through the lens (virtual target) can be visualized by the beam’s-eye monitor. By manipulating the vernier dials of the frame fixation adaptor, the virtual target is placed at the position that indicates the isocenter. A similar procedure is repeated at the gantry rotation angle of 0” (top position) so that the target point is 3-dimensionally placed at the isocenter. Coincidence of the viewpoint of the video camera and the x-ray source is confirmed periodically by using a small metal ball and x-ray film. The coincidence is not strictly necessary because a paticular point in the camera view field indicates the isocenter of the machine. Here, isocenter is defined based on monitor images of the metal ball taken at several different gantry and collimator an-
AND
RESULTS
Error due to the optical aberration Due to the optical aberration of the convex lens, some degree of displacement of the image of the virtual target might occur if the optical axis of the lens is not identical to that of the beam’s-eye monitor. The diameter of the lens is 18 mm and the distance from the lens to the center of the stereotactic arc is 230 mm. Therefore, the displacement angle (a) should be: CY < tan-’ (18/2/230) = 2.2” Although the virtual target is usually adjusted to the center of the lens field, we set it in a crucial condition to investigate the maximum displacement. The displacement was less than 0.3 mm in the set condition (Fig. 4) and undetectable when angle (Y was less than l”, a condition very easy to satisfy during the positioning. Localization and positioning A test apparatus (Fig. 5) was prepared for measuring the localization and positioning accuracy of the system. A tungsten ball fixed to the stereotactic base frame was used as a simulated target point. A stereotactic localization box for angiogram was attached to the base frame, and two orthogonal x-ray films were exposed for antero-posterior and left-right views of the phantom (Fig. 6). The coordinates of the simulated target and the 20 fiducial points on the indicator were read out from 4 sets of x-ray films, and the stereotactic coordinates of the simulated target point were calculated. Localization procedures with a CT scanner (Xvision/GX: Toshiba Inc., Tokyo, Japan) were also performed with 1 mm thickness and a spacing interval of the same value (Fig. 6b, c). The image field of view was 240 mm in diameter, with a 512 X 512 matrix. Four series of CT scans were also performed. To prove reproducibility, the frame and simulated target were removed and reattached to the CT couch for each series of scans. The couch was set at a different height from the floor for each series. The mean values and the standard deviations of the coordinates obtained by the x-ray and CT procedure are shown in Table 1. Here, the x, y, and z directions are the right-to-left, back-to-front, and foot-to-head directions for patients, respectively. Standard deviations of the target coordinates by x-ray localization were approximately
142
I. J. Radiation Oncology
l
Biology
l
Physics
Volume 40, Number 3. 1998
Fig. 3. The target pointer can be directed to the X-ray source in any possible combination
0.1 mm. On the other hand, by CT localization, standard deviations in z coordinate measurementswere 0.8 mm at maximum. Therefore, we used the average coordinates obtained from x-ray localizations for further measurements. Accuracy of the positioning The base frame was mounted on the fixation adapter at the end of the treatment couch in the neutral rotation position (perpendicular to the gantry’s rotation plane). The simulatedtarget was then placed to the isocenterof the linac by meansof the indicator and the beam’s-eye monitor, just as is done for the routine positioning of the patients. The
of gantry and coach rotation.
simulated target was hidden during the positioning procedure, so only the acquired coordinate of the target was referenced. After the procedure was accomplished,displacementsof the simulated targets from the isocenter were measured (Table 2). Displacements in the x and y directions were measuredand confirmed from the portal verification films, using the linac’s radiation beam taken at the top, bottom, right, and left positions of the gantry (Fig. 6d). The z displacementwasdefined from exposureswith two different collimator rotations of the machine. The simulated target was then removed from the base frame and reattached in a different position to repeat two
Table 1. X-ray and CT localization X-ray localization
CT localization
(mm)
Target
x
Y
z
1 2 3
103.1 -e .l 123.7 2 .l 60.4 ? .l
97.6 ? .l 84.9 k .l 127.3 k .l
95.5 ” .l 86.7 2 .l 73.9 ? .l
X
103.0 k .4 123.5 + .4 60.0 ? .4
(mm)
Y
z
97.3 -c .3 84.6 If: .3 126.9 2 .3
96.0 2 .8 87.3 ? .7 14.2 ? .4
Gantry-mountedvideo cameral
E. KUNIEDA
et al.
143
Head
Fig. 4. Bullseye images through the beam’s eye monitor when the gantry is in the upmost position. The image in the center demonstrates that the virtual target (the center of the bullseye) is at “right” position on the beam’s-eye monitor. The crossing white lines are electrically generated in the monitor display and indicate the isocenter. Four peripheral images show that the movements of the virtual target were minimal when the target pointer was inclined approximately 2” (a = 2) to the right, left, head, and foot direction in the patient’s supine position. The diameters of the circles are 5, 3, and 1 mm.
more series of x-ray localizations and positionings. A series of 3 measurements was repeated for each target. The maximum error of the repositioning of the simulated target during all series of measurements was less than 0.3 mm. DISCUSSION Localization and positioning accuracies are significant and challenging tasks in radiosurgery. The degree of accuracy required may differ from case to case. If, for example, we treat a small acoustic neurinoma, geometrical accuracies are very crucial. Acoustic neurinomas are tumors that are very often nonspherical in shape and ideally treated with multiple foci of beam concentration (multiple isocenter). The margin of the tumor is fairly demarcated and it is comparatively easy to determine the
target volume for irradiation. However, facial nerves usually locate along the margin of the tumors (12) and radiation damage of the nerves may cause nerve palsy, a devastating adverse side effect that should be avoided at all costs. Concentrated linac beams collimated by secondary collimators have a much steeper dose fall-off than that of gamma units (13). Thus, slight displacement of the positioning to the target might cause substantial changes Table 2. Root mean square of the target displacements (mm) Target
Ax AY AZ
1
2
3
0.10 0.12 0.15
0.08 0.12 0.26
0.18 0.16 0.26
I. J. Radiation
Oncology
0 Biology
l
Physics
Volume
40, Number
3, 1998
localization is more accurate than CT localization because the accuracy of the latter is limited by the pixel size and the thickness of the slice. Unlike the BRW frame, the Leksell stereotactic system does not utilize absolute (mechanical) coordinates as references. Nevertheless, the coordinates obtained by orthogonal films and those by CT scans almost coincide with each other. Furthermore, the measurements were based on the acquired coordinates of the simulated targets. The results indicated that the localizations were performed with sufficient reproducibility and that the positionings were satisfactorily accomplished. Many of the radiosurgery systems reported to date (7, 9, 10) utilize positioning methods that depend on wallmounted laser beams. The widths of these laser beams necessarily limit the accuracy of these systems. Although positioning should make laser beams ideally intersect at the isocenter, the tendency of the beams to lose their alignment during machine use necessitates repeated adjustments (17, 18). Lutz er al. (8) utilized a floor-mounted stereotactic frame support assembly. Instead of using laser beams, the target positioning in their system is accomplished using a simulated target mounted on a special floor stand placed on the turntable of the linac. The simulated target is replaced by the patient’s head during treatment. Although precise positioning accuracy can be expected, the procedure is rather complex. Fig. 5. Test apparatus for measuring the positioning accuracy of the system. A tungsten ball at the end of a metal bar was supported by a three-axis slide assembly attached to the stereotactic base frame.
Characteristics of our system The characteristics of our system can be described as follows.
in the irradiated dose of the area peripheral to the target volume. Moreover, when multiple isocenters are concerned, the distances among the focus points are crucial to the formation of the intended dose distribution. If the distances of the multiple isocenters become wider than intended, the dosage in their conjunctional areas might be lower than expected; if they become narrower, higher dosage areas might occur. The risks of facial neuropathy following radiosurgery increase with increasing dosage and the increasing length of the nerve in the radiation field (14). Unintended high-dose irradiation to the nerve or inadequate dose distributions due to errors of positioning may increase the incidence of nerve palsy. On the other hand, the positioning accuracy required for a single-focus irradiation for a metastatic tumor with a wider irradiation field is not as tight as that required for acoustic neurinoma. As noted in several reports (8, 15, 16), x-ray localizations produce quite precise target coordinates (most of the measurements in the reports were carried out using BRW stereotactic frames). These reports also indicate that x-ray
1. The positioning mechanism is independent of the wall or ceiling laser beams. The lens and the beam’s-eye monitor make it easy to place the arc center at the isocenter. The positions of the bullseye and the lens do not need to be strictly placed on the line between the x-ray source and the target point. Instead, with the help of their optical geometry, the direction of the target indicator is allowed to be roughly placed as long as the bullseye appears through the lens. 2. The image of the bullseye in the beam’s-eye monitor is magnified. Virtually any spatial resolution of the image can be obtained by changing the magnification factor of the lenses. 3. Minor rotations or displacements of the frame are negligible. The top plate of the couch bends downward by the weight of the patient’s head. Because our positioning procedure is dependent only on the geometrical relation between the gantry and the center of the arc system, it is practically unaffected by the minor rotation of the frame. 4. Positionings are possible when the couch rotation around the isocenter is not in a neutral position. We
Gantry-mounted
video camera
(4
l
E. KUNIEDA
et al.
(b)
(4 Fig. 6. The coordinates of the simulated targets (tungsten ball) were obtained from orthogonal x-ray films (a). An axial CT image (b) and a scoutgraphy of the test apparatus (c) are shown. After the positioning, portal films of the simulated target were exposed and the differences of the center of the circular field (d = 17 mm) and the simulated target (d = 4 mm) were measured (d).
can confirm and adjust the position of the target in each rotation of the treatment couch during multiarc converging irradiation. This capability of our system is quite effective, especially when high positioning accuracy is required. The center of the couch rotation around the isocenter tends not to coincide with the
center of the gantry rotation. These centers are very difficult to adjust after the gantry and the couch have been installed. In most of the reported positioning methods that use wall-mounted laser beams, it is difficult to position the target when the couch is not in the neutral position.
146
I. J. Radiation Oncology
l
Biology 0 Physics
CONCLUSIONS Although the major source of errors in radiosurgery positioning comes from the localization of the target (19) a submillimeter order of accuracy is often required for smallfield irradiations. The Keio University stereotactic radiosurgery systemwas designed to be independent of wall-mounted laser beams.
Volume 40, Number 3, 1998
Instead, an arc system and gantry-mounted video camera were utilized for the positioning of the target point. Although there was a fear that high-energy photons might degradethe functions of the semiconductorcircuits or CCD sensor of the video camera, we have not encountered any trouble related to the CCD video system since the operation of this system was started in April, 1994.
REFERENCES 1. Larsson, B.; Liden, K.; Sarby, B. Irradiation of small structures through the intact skull. Acta Radio]. Ther. Phys. Biol. 13:512-534; 1974. 2. Leksell, L. Sterotaxic radiosurgery in trigeminal neuralgia. Acta Chir. Stand. 137:311-314; 1971. 3. Flickinger, J. C.; Lunsford, L. D.; Coffey, R. J.; Linskey, M. E.; Bissonette, D. J.; Maitz, -4. H.; Kondziolka, D. Radiosurgery of acoustic neurinomas. Cancer 67:345-353; 1991. 4. Loeffler, J. S.; Kooy, H. M.; Wen, P. Y.; Fine, H. A.; Cheng, C. W.; Mannarino, E. G.; Tsai, J. S.; Alexander, E. D. The treatment of recurrent brain metastases with stereotactic radiosurgery. J. Clin. Oncol. 8:576-582; 1990. 5. Steiner, L.; Lindquist, C.; Adler, J. R.; Tomer, J. C.; Alves, W.; Steiner, M. Clinical outcome of radiosurgery for cerebral arteriovenous malformations. J. Neurosurg. 77:1-8; 1992. 6. Colombo, F.; Benedetti, A.; Pozza, F.; Avanzo, R. C.; Marchetti, C.; Chierego, G.; Zanardo, A. External stereotactic irradiation by linear accelerator. Neurosurgery 16:154-160; 1985. 7. Hartmann, G. Cerebral radiation surgery using moving field irradiation at a linear accelerator facility. Int. J. Radiat. Oncol. Biol. Phys. 11:1185-1192; 1985. 8. Lutz, W.; Winston, K. R.; Maleki, N. A system for stereotactic radiosurgery with a linear accelerator. Int. J. Radiat. Oncol. Biol. Phys. 14:373-381; 1988. 9. Podgorsak, E. B.; Olivier, A.; Pla, M.; Lefebvre, P. Y.; Hazel, J. Dynamic stereotactic radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 14: 115-126; 1988. 10. Tokuuye, K.; Akine, Y.; Tokita, N., et al. Linac-based smallfield radiotherapy for brain tumors. Radiother. Oncol. 27:5558; 1993. Il. Gibbs, F. J.; Buechler, D.; Leavitt, D. D.; Moeller, J. H. Measurement of mechanical accuracy of isocenter in conventional linear-accelerator-based radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 25:117-122; 1993.
12. Koos, W. T.; Matula, C.; Levy, D.; Kitz, K. Microsurgery vs. radiosurgery in the treatment of small acoustic neurinomas. Acta Neurochir. Suppl. Wien 63:73-80; 1995. 13. Podgorsak, E. B.; Pike, G. B.; Olivier, A.; Pla, M.; Souhami, L. Radiosurgery with high energy photon beams: a comparison among techniques. Int. J. Radiat. Oncol. Biol. Phys. 16: 857-865; 1989. 14. Linskey, M. E.; Flickinger, J. C.; Lunsford, L. D. Cranial nerve length predicts the risk of delayed facial and trigeminal neuropathies after acoustic tumor stereotactic radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 25:227-233; 1993. 15. Serago, C. F.; Lewin, A. A.; Houdek, P. V.; Gonzalez, A. S.; Hartmann, G. H.; Abitbol, A. A.; Schwade, J. G. Stereotactic target point verification of an Xray and CT localizer. Int. J. Radiat. Oncol. Biol. Phys. 20:517-523; 1991. 16. 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. 17. Heifetz, M. D.; Rosemark, P. J.; Wexler, M. C.; Greenberg, S. H.; Thompson, R. W. Rapid method for determination of isocenter of radiation gantry and alignment of laser beams for stereotactic radiosurgery. Stereotact. Funct. Neurosurg. 53: 46-48; 1989. 18. Serago, C. F.; Lewin, A. A.; Houdek, P. V.; Gonzalez, A. S.; Schwade, J. G.; Abitbol, A.; Martial, V. V. Radiosurgery target point alignment errors detected with portal film verification. Int. J. Radiat. Oncol. Biol. Phys. 24:777-780; 1992. 19. Freeman, C. R.; Souhami, L.; Caron, J. L.; Villemure, J. G.; Olivier, A.; Montes, J.; Farmer, J. P.; Podgorsak, E. B. Stereotactic external beam irradiation in previously untreated brain tumors in children and adolescents. Med. Pediatr. Oncol. 22:173-180; 1994.
Biol.
Phya., Vol. 40. No. 3. pp. 739-746. IYYX Copyright 0 199X Elsrvirr Science Inc. Printed in the USA. All right\ reserved 0360-3016/YX $I’).00 + .OO
PI1 SO360-3016(97)00844-4
ELSEVIER
l
Oncology
Physics Contribution NEW SYSTEM
FOR LINEAR ACCELERATOR GANTRY-MOUNTED VIDEO
ETSUO KUNIEDA, M.D.,* MASAYUKI KITAMURA, M.D.,* TAKAYUKI OHIRA, M.D.,’ KOUICHI OGAWA, PH.D.,*
KAYOKO NAKAMURA
PH.D.* AND ATSUSHI
RADIOSURGERY CAMERA OSAMU
A
M.D.,* M.D.,*
KAWAGUCHI,
YUTAKA ANDO, KUBO,
WITH
M.D.*
*Departmentsof Radiology,and ‘Neurosurgery,Schoolof Medicine, Keio University, Tokyo, Japan;and“Departmentof Electrical Engineering,Collegeof Engineering,HoseiUniversity. Tokyo, Japan Purpose: We developed a positioning method that does not depend on the positioning mechanism originally annexed to the linac and investigated the positioning errors of the system. Methods and Materials: A small video camera was placed at a location optically identical to the linac x-ray source. A target pointer comprising a convex lens and bull’s eye was attached to the arc of the Leksell stereotactic system so that the lens would form a virtual image of the bull’s eye (virtual target) at the position of the center of the arc. The linac gantry and target pointer were placed at the side and top to adjust the arc center to the isocenter by referring the virtual target. Coincidence of the target and the isocenter could be confirmed in any combination of the couch and gantry rotation. In order to evaluate the accuracy of the positioning, a tungsten ball was attached to the stereotactic frame as a simulated target, which was repeatedly localized and repositioned to estimate the magnitude of the error. The center of the circular field defined by the collimator was marked on the film. Results: The differences between the marked centers of the circular field and the centers of the shadow of the simulated target were less than 0.3 mm. 0 1998 Elsevier Science Inc. Stereotactic
radiosurgery,
Linear
accelerator,
Video camera, Beam&eye
INTRODUCTION
monitoring.
on laser positioners originally annexed to the machine for conventional radiotherapy, and we report our examination of the accuracy of this positioning method.
Although the gamma unit produced by Leksell (1, 2) has been applied for more than 2 decades and has demonstrated the clinical utility of radiosurgery (3-5) there has been increasing interest in carrying out stereotactic radiosurgery using x-ray beams produced by linear accelerators, or linacs (6-10). In addition to the quantitative advantages of linac-based radiosurgery over gamma units, such as low initial introductory cost and the absence of any need for source replacement, the linac’s increased source-collimator distance (8) reduces limitation in the size of the radiation field. On the other hand, most of the medical linacs currently in use were not originally designedto be applied to radiosurgery: the accuracy required for this purpose exceeds the level that the vendors of these therapy machineshave assumedtheir products to have. We describe in this paper a new radiosurgery system that employs a positioning method that does not depend
The linear accelerator and auxiliary devices A 6MV linear accelerator (MLlSMV: Mitsubishi Electries Corp., Tokyo, Japan) was used to produce the x-ray beam for stereotactic radiosurgery. We utilized secondary collimators with 5 mm to 30 mm-diameter openings projected to the isocenter of the machine. The height of the collimator is 110 mm and its isocenter-side end is positioned 320 mm away from the isocenter of the machine. During treatment, the base frame securing the patient’s head is fitted to the treatment couch by an adaptor equipped with three orthogonal drives and verniers.
Presented at First Korea-Japan Joint Meeting on Medical Physits. Seoul,Korea. September19. 1996. Reprint requeststo: EtsuoKunieda,M.D.. Departmentof Ra-
diology. Keio University, 35 Shinanomachi, Shinjuku.Tokyo 160 Japan. Acceptedfor publication25 September1997.
METHODS
739
I. J. RadiationOncology l Biology l Physics
140 X-ray
'\
source
Volume 40, Number 3, 1998
CCD video camera and the light source
Stereotactic
Target Stereotactic ar -
-..-*.. .._ -.._--. ... .. ...I----v
/-
arc
’
pointer
Convex lens Bull's eye
“Virtual target” arc center
at the c
Fig. 1. A schematic illustration of the positioning method. A small CCD video camera (beam’s-eye monitor) is mounted at a point optically identical to the linac x-ray source. The center of the arc is adjusted to the gantry isocenter under the guidance of the “virtual target” produced by a convex lens and a bullseye (target pointer) mounted on the stereotactic XC.
Beam ‘s-eye monitor In the gantry of a linac, a small CCD (charge-coupled device) video camera is mounted at the position where the source of the light field is usually placed. The light source and the video camera are alternately guided into the exact position when each of them is needed. The video camera is located at the position optically identical to the linac x-ray source, enabling real-time observation of the margin and the center of the irradiation field. Because the video camera displays the irradiation field as if viewed from the eye point of the x-ray beam source, we refer to it as a beam’s-eye monitor (Fig. 1). The video signal from the camera is digitized by means of the video-capture function of a personal computer (Macintosh 7100AV: Apple Computer Inc., Cupertino, CA). The matrix of the image is 640 X 480, with a pixel size of 0.14 X 0.14 mm. The images of the beam’s-eye monitor can be seen as motion images or static images at any moment. Arc system and target pointer Instead of depending on the beam’s alignment, we adopted a positioning method that is independent of the
room laser beams. The schematic shown in Fig. 1 gives a simple overview of our method. A positioning device, a target pointer, is attached to the arc of the stereotactic system. The arc is a part of the Leksell stereotactic system (Leksell Stereotactic system Type G, Elekta Instruments, AB, Stockholm, Sweden) and is used primarily to hold a needle for biopsy or aspiration of intracranial fluid. The target pointer slides along the arc with its axis kept perpendicular to the arc. A convex lens and bullseye are mounted on the target pointer in the same optical alignment along its long axis. Because the head of the patient is kept in the base frame, it is impossible to place into the machine isocenter an object that would indicate the arc center. Nevertheless, the lens makes a virtual image of the bullseye, a virtual target, at the position of the arc center, even though the actual position of the bullseye is far way from the center (Fig. 2). The target pointer pivots on the center of the arc, and we can turn the target pointer toward any direction from which the accelerator’s beam might come (Fig. 3). The stereotactic arc is mounted on the base frame so that the center of the arc can be matched with the intended target
Gantry-mountedvideo camera0
CCD camera L,
( f= f,)
l/a-l/b=l/ft
E. KUNIEDA
741
et al.
gles in a similar way to the definition of the mechanical isocenter. Because the largest displacements of the gantry are at the top and bottom positions of the gantry rotation (1 l), only the adjustment parallel to the gantry rotation plane is performed at the gantry’s top position. If collimators of less than 10 mm are used, these positioning procedures are repeatedly carried out in each couch rotation for an arc irradiation.
l/D+l/d=l/fc MEASUREMENTS
L+I=SAD
I I I
---
+L-_--d
Fig. 2. Optical geometry of the CCD camera and the target pointer when the arc center coincides with the isocenter of the gantry rotation. L, and L,., the lenses of the target pointer and the CCD camera, have focal distances off, and f,, respectively. The bullseye (B) and its virtual image (B’) are indicated in the figure. SAD = Source axis distance. point by referring to the Cartesian coordinates obtained by the localization procedure. The target point is roughly positioned at the isocenter of the linac using the conventional laser beam positioner in advance to obtain more precise positioning. Next, the linac gantry and the target pointer are positioned at 90” or 270” (side position), so that the image of the bullseye through the lens (virtual target) can be visualized by the beam’s-eye monitor. By manipulating the vernier dials of the frame fixation adaptor, the virtual target is placed at the position that indicates the isocenter. A similar procedure is repeated at the gantry rotation angle of 0” (top position) so that the target point is 3-dimensionally placed at the isocenter. Coincidence of the viewpoint of the video camera and the x-ray source is confirmed periodically by using a small metal ball and x-ray film. The coincidence is not strictly necessary because a paticular point in the camera view field indicates the isocenter of the machine. Here, isocenter is defined based on monitor images of the metal ball taken at several different gantry and collimator an-
AND
RESULTS
Error due to the optical aberration Due to the optical aberration of the convex lens, some degree of displacement of the image of the virtual target might occur if the optical axis of the lens is not identical to that of the beam’s-eye monitor. The diameter of the lens is 18 mm and the distance from the lens to the center of the stereotactic arc is 230 mm. Therefore, the displacement angle (a) should be: CY < tan-’ (18/2/230) = 2.2” Although the virtual target is usually adjusted to the center of the lens field, we set it in a crucial condition to investigate the maximum displacement. The displacement was less than 0.3 mm in the set condition (Fig. 4) and undetectable when angle (Y was less than l”, a condition very easy to satisfy during the positioning. Localization and positioning A test apparatus (Fig. 5) was prepared for measuring the localization and positioning accuracy of the system. A tungsten ball fixed to the stereotactic base frame was used as a simulated target point. A stereotactic localization box for angiogram was attached to the base frame, and two orthogonal x-ray films were exposed for antero-posterior and left-right views of the phantom (Fig. 6). The coordinates of the simulated target and the 20 fiducial points on the indicator were read out from 4 sets of x-ray films, and the stereotactic coordinates of the simulated target point were calculated. Localization procedures with a CT scanner (Xvision/GX: Toshiba Inc., Tokyo, Japan) were also performed with 1 mm thickness and a spacing interval of the same value (Fig. 6b, c). The image field of view was 240 mm in diameter, with a 512 X 512 matrix. Four series of CT scans were also performed. To prove reproducibility, the frame and simulated target were removed and reattached to the CT couch for each series of scans. The couch was set at a different height from the floor for each series. The mean values and the standard deviations of the coordinates obtained by the x-ray and CT procedure are shown in Table 1. Here, the x, y, and z directions are the right-to-left, back-to-front, and foot-to-head directions for patients, respectively. Standard deviations of the target coordinates by x-ray localization were approximately
142
I. J. Radiation Oncology
l
Biology
l
Physics
Volume 40, Number 3. 1998
Fig. 3. The target pointer can be directed to the X-ray source in any possible combination
0.1 mm. On the other hand, by CT localization, standard deviations in z coordinate measurementswere 0.8 mm at maximum. Therefore, we used the average coordinates obtained from x-ray localizations for further measurements. Accuracy of the positioning The base frame was mounted on the fixation adapter at the end of the treatment couch in the neutral rotation position (perpendicular to the gantry’s rotation plane). The simulatedtarget was then placed to the isocenterof the linac by meansof the indicator and the beam’s-eye monitor, just as is done for the routine positioning of the patients. The
of gantry and coach rotation.
simulated target was hidden during the positioning procedure, so only the acquired coordinate of the target was referenced. After the procedure was accomplished,displacementsof the simulated targets from the isocenter were measured (Table 2). Displacements in the x and y directions were measuredand confirmed from the portal verification films, using the linac’s radiation beam taken at the top, bottom, right, and left positions of the gantry (Fig. 6d). The z displacementwasdefined from exposureswith two different collimator rotations of the machine. The simulated target was then removed from the base frame and reattached in a different position to repeat two
Table 1. X-ray and CT localization X-ray localization
CT localization
(mm)
Target
x
Y
z
1 2 3
103.1 -e .l 123.7 2 .l 60.4 ? .l
97.6 ? .l 84.9 k .l 127.3 k .l
95.5 ” .l 86.7 2 .l 73.9 ? .l
X
103.0 k .4 123.5 + .4 60.0 ? .4
(mm)
Y
z
97.3 -c .3 84.6 If: .3 126.9 2 .3
96.0 2 .8 87.3 ? .7 14.2 ? .4
Gantry-mountedvideo cameral
E. KUNIEDA
et al.
143
Head
Fig. 4. Bullseye images through the beam’s eye monitor when the gantry is in the upmost position. The image in the center demonstrates that the virtual target (the center of the bullseye) is at “right” position on the beam’s-eye monitor. The crossing white lines are electrically generated in the monitor display and indicate the isocenter. Four peripheral images show that the movements of the virtual target were minimal when the target pointer was inclined approximately 2” (a = 2) to the right, left, head, and foot direction in the patient’s supine position. The diameters of the circles are 5, 3, and 1 mm.
more series of x-ray localizations and positionings. A series of 3 measurements was repeated for each target. The maximum error of the repositioning of the simulated target during all series of measurements was less than 0.3 mm. DISCUSSION Localization and positioning accuracies are significant and challenging tasks in radiosurgery. The degree of accuracy required may differ from case to case. If, for example, we treat a small acoustic neurinoma, geometrical accuracies are very crucial. Acoustic neurinomas are tumors that are very often nonspherical in shape and ideally treated with multiple foci of beam concentration (multiple isocenter). The margin of the tumor is fairly demarcated and it is comparatively easy to determine the
target volume for irradiation. However, facial nerves usually locate along the margin of the tumors (12) and radiation damage of the nerves may cause nerve palsy, a devastating adverse side effect that should be avoided at all costs. Concentrated linac beams collimated by secondary collimators have a much steeper dose fall-off than that of gamma units (13). Thus, slight displacement of the positioning to the target might cause substantial changes Table 2. Root mean square of the target displacements (mm) Target
Ax AY AZ
1
2
3
0.10 0.12 0.15
0.08 0.12 0.26
0.18 0.16 0.26
I. J. Radiation
Oncology
0 Biology
l
Physics
Volume
40, Number
3, 1998
localization is more accurate than CT localization because the accuracy of the latter is limited by the pixel size and the thickness of the slice. Unlike the BRW frame, the Leksell stereotactic system does not utilize absolute (mechanical) coordinates as references. Nevertheless, the coordinates obtained by orthogonal films and those by CT scans almost coincide with each other. Furthermore, the measurements were based on the acquired coordinates of the simulated targets. The results indicated that the localizations were performed with sufficient reproducibility and that the positionings were satisfactorily accomplished. Many of the radiosurgery systems reported to date (7, 9, 10) utilize positioning methods that depend on wallmounted laser beams. The widths of these laser beams necessarily limit the accuracy of these systems. Although positioning should make laser beams ideally intersect at the isocenter, the tendency of the beams to lose their alignment during machine use necessitates repeated adjustments (17, 18). Lutz er al. (8) utilized a floor-mounted stereotactic frame support assembly. Instead of using laser beams, the target positioning in their system is accomplished using a simulated target mounted on a special floor stand placed on the turntable of the linac. The simulated target is replaced by the patient’s head during treatment. Although precise positioning accuracy can be expected, the procedure is rather complex. Fig. 5. Test apparatus for measuring the positioning accuracy of the system. A tungsten ball at the end of a metal bar was supported by a three-axis slide assembly attached to the stereotactic base frame.
Characteristics of our system The characteristics of our system can be described as follows.
in the irradiated dose of the area peripheral to the target volume. Moreover, when multiple isocenters are concerned, the distances among the focus points are crucial to the formation of the intended dose distribution. If the distances of the multiple isocenters become wider than intended, the dosage in their conjunctional areas might be lower than expected; if they become narrower, higher dosage areas might occur. The risks of facial neuropathy following radiosurgery increase with increasing dosage and the increasing length of the nerve in the radiation field (14). Unintended high-dose irradiation to the nerve or inadequate dose distributions due to errors of positioning may increase the incidence of nerve palsy. On the other hand, the positioning accuracy required for a single-focus irradiation for a metastatic tumor with a wider irradiation field is not as tight as that required for acoustic neurinoma. As noted in several reports (8, 15, 16), x-ray localizations produce quite precise target coordinates (most of the measurements in the reports were carried out using BRW stereotactic frames). These reports also indicate that x-ray
1. The positioning mechanism is independent of the wall or ceiling laser beams. The lens and the beam’s-eye monitor make it easy to place the arc center at the isocenter. The positions of the bullseye and the lens do not need to be strictly placed on the line between the x-ray source and the target point. Instead, with the help of their optical geometry, the direction of the target indicator is allowed to be roughly placed as long as the bullseye appears through the lens. 2. The image of the bullseye in the beam’s-eye monitor is magnified. Virtually any spatial resolution of the image can be obtained by changing the magnification factor of the lenses. 3. Minor rotations or displacements of the frame are negligible. The top plate of the couch bends downward by the weight of the patient’s head. Because our positioning procedure is dependent only on the geometrical relation between the gantry and the center of the arc system, it is practically unaffected by the minor rotation of the frame. 4. Positionings are possible when the couch rotation around the isocenter is not in a neutral position. We
Gantry-mounted
video camera
(4
l
E. KUNIEDA
et al.
(b)
(4 Fig. 6. The coordinates of the simulated targets (tungsten ball) were obtained from orthogonal x-ray films (a). An axial CT image (b) and a scoutgraphy of the test apparatus (c) are shown. After the positioning, portal films of the simulated target were exposed and the differences of the center of the circular field (d = 17 mm) and the simulated target (d = 4 mm) were measured (d).
can confirm and adjust the position of the target in each rotation of the treatment couch during multiarc converging irradiation. This capability of our system is quite effective, especially when high positioning accuracy is required. The center of the couch rotation around the isocenter tends not to coincide with the
center of the gantry rotation. These centers are very difficult to adjust after the gantry and the couch have been installed. In most of the reported positioning methods that use wall-mounted laser beams, it is difficult to position the target when the couch is not in the neutral position.
146
I. J. Radiation Oncology
l
Biology 0 Physics
CONCLUSIONS Although the major source of errors in radiosurgery positioning comes from the localization of the target (19) a submillimeter order of accuracy is often required for smallfield irradiations. The Keio University stereotactic radiosurgery systemwas designed to be independent of wall-mounted laser beams.
Volume 40, Number 3, 1998
Instead, an arc system and gantry-mounted video camera were utilized for the positioning of the target point. Although there was a fear that high-energy photons might degradethe functions of the semiconductorcircuits or CCD sensor of the video camera, we have not encountered any trouble related to the CCD video system since the operation of this system was started in April, 1994.
REFERENCES 1. Larsson, B.; Liden, K.; Sarby, B. Irradiation of small structures through the intact skull. Acta Radio]. Ther. Phys. Biol. 13:512-534; 1974. 2. Leksell, L. Sterotaxic radiosurgery in trigeminal neuralgia. Acta Chir. Stand. 137:311-314; 1971. 3. Flickinger, J. C.; Lunsford, L. D.; Coffey, R. J.; Linskey, M. E.; Bissonette, D. J.; Maitz, -4. H.; Kondziolka, D. Radiosurgery of acoustic neurinomas. Cancer 67:345-353; 1991. 4. Loeffler, J. S.; Kooy, H. M.; Wen, P. Y.; Fine, H. A.; Cheng, C. W.; Mannarino, E. G.; Tsai, J. S.; Alexander, E. D. The treatment of recurrent brain metastases with stereotactic radiosurgery. J. Clin. Oncol. 8:576-582; 1990. 5. Steiner, L.; Lindquist, C.; Adler, J. R.; Tomer, J. C.; Alves, W.; Steiner, M. Clinical outcome of radiosurgery for cerebral arteriovenous malformations. J. Neurosurg. 77:1-8; 1992. 6. Colombo, F.; Benedetti, A.; Pozza, F.; Avanzo, R. C.; Marchetti, C.; Chierego, G.; Zanardo, A. External stereotactic irradiation by linear accelerator. Neurosurgery 16:154-160; 1985. 7. Hartmann, G. Cerebral radiation surgery using moving field irradiation at a linear accelerator facility. Int. J. Radiat. Oncol. Biol. Phys. 11:1185-1192; 1985. 8. Lutz, W.; Winston, K. R.; Maleki, N. A system for stereotactic radiosurgery with a linear accelerator. Int. J. Radiat. Oncol. Biol. Phys. 14:373-381; 1988. 9. Podgorsak, E. B.; Olivier, A.; Pla, M.; Lefebvre, P. Y.; Hazel, J. Dynamic stereotactic radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 14: 115-126; 1988. 10. Tokuuye, K.; Akine, Y.; Tokita, N., et al. Linac-based smallfield radiotherapy for brain tumors. Radiother. Oncol. 27:5558; 1993. Il. Gibbs, F. J.; Buechler, D.; Leavitt, D. D.; Moeller, J. H. Measurement of mechanical accuracy of isocenter in conventional linear-accelerator-based radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 25:117-122; 1993.
12. Koos, W. T.; Matula, C.; Levy, D.; Kitz, K. Microsurgery vs. radiosurgery in the treatment of small acoustic neurinomas. Acta Neurochir. Suppl. Wien 63:73-80; 1995. 13. Podgorsak, E. B.; Pike, G. B.; Olivier, A.; Pla, M.; Souhami, L. Radiosurgery with high energy photon beams: a comparison among techniques. Int. J. Radiat. Oncol. Biol. Phys. 16: 857-865; 1989. 14. Linskey, M. E.; Flickinger, J. C.; Lunsford, L. D. Cranial nerve length predicts the risk of delayed facial and trigeminal neuropathies after acoustic tumor stereotactic radiosurgery. Int. J. Radiat. Oncol. Biol. Phys. 25:227-233; 1993. 15. Serago, C. F.; Lewin, A. A.; Houdek, P. V.; Gonzalez, A. S.; Hartmann, G. H.; Abitbol, A. A.; Schwade, J. G. Stereotactic target point verification of an Xray and CT localizer. Int. J. Radiat. Oncol. Biol. Phys. 20:517-523; 1991. 16. 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. 17. Heifetz, M. D.; Rosemark, P. J.; Wexler, M. C.; Greenberg, S. H.; Thompson, R. W. Rapid method for determination of isocenter of radiation gantry and alignment of laser beams for stereotactic radiosurgery. Stereotact. Funct. Neurosurg. 53: 46-48; 1989. 18. Serago, C. F.; Lewin, A. A.; Houdek, P. V.; Gonzalez, A. S.; Schwade, J. G.; Abitbol, A.; Martial, V. V. Radiosurgery target point alignment errors detected with portal film verification. Int. J. Radiat. Oncol. Biol. Phys. 24:777-780; 1992. 19. Freeman, C. R.; Souhami, L.; Caron, J. L.; Villemure, J. G.; Olivier, A.; Montes, J.; Farmer, J. P.; Podgorsak, E. B. Stereotactic external beam irradiation in previously untreated brain tumors in children and adolescents. Med. Pediatr. Oncol. 22:173-180; 1994.