Ini J Radratron Oncology BIO/ Phvs Vol. Printed in the US A. All nghls reserved.

I’), pp.

1059-1065

??Technical Innovations

0360.3016/90 $3.00 + .OO Copynght G 1990 Pergamon Press plc

and Notes

VIRTUAL SIMULATION IN THE CLINICAL SETTING: SOME PRACTICAL CONSIDERATIONS GEORGE W. SHEROUSE, M.S.,’ J. DANIEL BOURLAND, PHD.,’ KEVIN REYNOLDS, B.S.,’ HARRIS L. MCMURRY, M.S.,’ THOMAS P. MITCHELL, M.S.2 AND EDWARD L. CHANEY, PH.D.’ ‘Department of Radiation Oncology, University of North Carolina, Chapel Hill, NC 27599-75 12; and ‘Department of Radiation Oncology, University of Florida. Gainesville, FL 326 10 Virtual simulation departs from normal practice by replacing conventional treatment simulation with 3-dimensional image data and computer software. Implementation of virtual simulation requires the ability to transfer the planned treatment geometry from the computer to the treatment room in a way which is accurate, reproducible, and efficient enough for routine use. We have separated this process into: (a) immobilization of the patient; (b) establishment and alignment of a practical coordinate system for the patient/couch system; and (c) setup of the patient/couch system and the treatment machine “by the numbers” to achieve the desired beam orientation. The first aspect has been addressed by the use of hemi- or full-body foam casts, the second by use of an alignment jig on the treatment couch, and the third with the aid of a patient coordinate system referenced to easily located landmarks. Phantom studies and clinical practice have shown these techniques to be practical and effective within reasonable clinical bounds. Treatment planning, Computer-aided treatment planning.

treatment design, Virtual simulation, Immobilization,

INTRODUCTION

Coordinate systems for

( 1) Rigidization: The planning image dataset represents a snapshot of the patient in space and time. Typically the assumption is made in the treatment planning process that the patient is a rigid body corresponding to that snapshot. A mechanism is required to assure that the actual, mobile patient will match the planning image within acceptable tolerances at the time of therapy. This is usually accomplished through the use of one or more forms of immobilization. (2) Coordinate transfer: The geometric relationship between patient and beams in a computer plan are expressed in terms of one or more coordinate systems which help define the computer models of the patient, table, beams, and treatment machine. Plan coordinates must be easily transferrable for the patient’s initial treatment. (3) Cost-effectiveness: The methods and materials used for achieving patient immobilization and coordinate transfer must be cost-effective, not just in a monetary sense, but also in terms of personnel time and impact on the patient.

In recent years there has been significant interest in the development of radiotherapy treatment planning systems that not only perform dose calculations but also provide a superior alternative to conventional simulation for accurately planning complex treatment setups. These new systems (5, 6, 7, 10, 11, 15) have in common the use of 3-dimensional image datasets to model the patient, and programs for planning the treatment geometry. The latter functionality is needed to obtain the highly accurate treatment geometry required for experimental modalities, such as particle beams (2) and stereotactic implants (1,4, 17), and to support equally demanding applications in photon therapy, including radiosurgery (8, 9) and investigations into innovative treatment setups (3, 14, 15). Implementing better treatment setups in a way that preserves accuracy raises some important practical problems. This paper discusses our approachs to handling problems that have arisen in association with use of the Virtual Simulator ( 11, 14, 15), our software analog of the conventional simulator. To reproduce a computer generated treatment setup accurately, at least the following three deliverability criteria must be met:

Satisfaction of these criteria has been addressed differently for the various modalities mentioned earlier. In stereotactic implants and photon teletherapy for brain lesions, for instance, it can reasonably be assumed that the

Reprint requests to: George Sherouse, M.S.

Accepted 1059

for publication

26 April 1990.

1060

1. J. Radiation Oncology 0 Biology 0 Physics

skull and its contents are rigid. The coordinate transfer criterion is met using very aggressive mechanical means (4, 9). In general, these methods satisfy the cost-effectiveness criterion since alternative therapies, if any, are more expensive. For particle beam therapy. typical approaches to satisfying the rigidization and coordinate transfer criteria have involved elaborate patient molds and/or timeconsuming iterative patient positioning at the time of therapy ( 18). The cost-effectiveness criterion is not an issue. In cases where particle beams are preferred, there usually is no alternative treatment, and since this therapy is considered experimental, cost is usually of parenthetical concern. In contrast to computer planning methods for more exotic modalities, virtual simulation is intended as a method for improving conventional photon teletherapy. Typically, the acceptable tolerances for stereotaxis and particle therapy are much more demanding than for more conventional teletherapy. Also. from a practical point of view, no significant additional cost or technologist effort per patient treatment can be borne in the average radiotherapy department. Our approach to satisfying the deliverability criteria, therefore, represents a compromise between clinically acceptable delivery tolerances and costeffectiveness in the routine clinical setting. METHODS

AND

MATERIALS

We have investigated the implementation of virtual simulations using a three-part approach: (a) expressing setup instructions in terms of a practical. patient-related coordinate system: (b) improving and extending common immobilization techniques to provide hemi- and wholebody immobilization; and (c) fabricating custom immobilization casts in a simple frame that repositions the patient/cast system on the treatment table in a way that allows accurate coordinate transfer.

October 1990. Volume 19. Number 4

Fig. I. Patient coordinate system. The positive x-axis is towards the patient’s left, the positive y-axis is towards the patient’s anterior surface. and the positive z-axis is towards the patient’s feet.

ward to measure the height of isocenter above tabletop using the lateral lasers and a ruler or custom scale. The choice of a “floating” z origin which is not attached to the tabletop allows flexibility in repositioning the patient/ cast system longitudinally on the treatment couch if necessary.

Iumwbilizatiorl One approach under investigation for patient immobilization is to create a custom cast with the patient in the most comfortable and natural position consistent with treatment objectives. In general, this procedure involves building polyurethane foam casts which are contoured to prominent anatomic features. For example, for a pelvis or lung treatment we construct a foam mold which extends from below the knees to above the shoulders, and comes up high on both sides of the patient. The knees would typically be elevated to create a natural cradle for the patient’s buttocks. The combination of this cradle with an

Coordinate systems We use a coordinate system for the couch/cast/patient system which is based on traditional radiotherapy practice, and which is readily accepted in the clinical setting. Figure 1 illustrates this coordinate system. The x- and y-axes are chosen to be in the transverse plane. This is consistent with the convention used by most traditional 2D planning systems. The system is right-handed. with the z-axis pointing toward the foot of the couch. The most important aspect of the coordinate system, shown in Figure 2, is the selection of the origin. The x origin is at the center of the narrow dimension of the table top. The y origin is the top surface of either the table or the immobilization substrate, depending on which surface is more convenient to locate. The z origin is patient-specific. In some cases an anatomic landmark such as the supersternal notch is appropriate; in others a landmark on the patient cast is used. The origin is easy to locate in all three dimensions by the treatment technologist. Furthermore, it is straightfor-

Y

X

x=

0 at table center

Y = 0 at tabletop Fig. 2. Origin of patient coordinate system. The x-origin is positioned at the center of the lateral dimension of the tabletop, and the y-origin is at the effective tabletop. The effective tabletop is either the actual tabletop or the anterior surface of the styrofoam substrate of the immobilization cast, whichever is more clinically convenient to use. The z-origin is custom selected for each patient and is either a reliable anatomic landmark, such as the supersternal notch, or a point on the immobilization cast.

1061

Virtual simulation in the clinical setting 0 G. W. SHEROUSEet al.

impression of the shoulders provides good longitudinal stability. Rotational stability is accomplished by allowing the foam to fill in the spaces between the legs, under the arms, and along the patient’s sides. Figure 3 shows a patient immobilized for a lung treatment. Foam molds (illustrated in Fig. 4) are made of a twocomponent polyurethane foam which is mixed in liquid form and then poured inside a large plastic bag. A a” sytrofoam insulation board is used as a substrate. The substrate lends structural integrity to the cast and plays a key role in coordinate transfer. As the polyurethane foam rises and cures, it bonds to the Styrofoam board yielding a one-piece foam unit. The plastic bag functions as a container and as a protective barrier between the patient and the foam. Some anatomic sites are more difficult to immobilize with foam because they lack prominent features for the foam to cradle or mold around. Limbs. the pelvis, and the back of the head are good examples. In cases where foam casts are suitable, we augment the casts with thermal plastic molds which can be attached to the foam mold. In particular, for heads we connect the thermal plastic to the foam without using fasteners or connectors which might create CT artifacts and which require additional attention from the technologist. Figure 5 shows a form which is placed next to the patient’s head before pouring the liquid polyurethane. The forms create notches in the

Fig. 3. Patient immobilized for a lung treatment. Note position of knees and generous use of foam to fill in spaces between the legs, around the shoulders and armpits, and on the sides of the patient.

m a m m

polyurethane foam base CT/treatment wood frame

foam table

Fig. 4. Cross-section of foam mold showing polystyrene substrate, polurethane foam that forms a mold around the patient, and wood frame for coordinate transfer jig.

sides of the foam cast supporting the head. A thermal plastic face mask can then be molded in such a way that it snaps into the notches. If done with care, this technique results in a foam-cast/thermal-plastic mask combination which securely surrounds the head and accurately reproduces the treatment position. At the same time the mask can be easily placed or removed. Its easy removal is an important safety issue for patients who may become ill during scanning or treatment. The construction of immobilization casts as described above requires careful attention to ensure a custom, reproducible fit. For some patients, particularly large patients with abdominal or pelvic tumors. mobile and/or pliable tissue mass is an obstacle to reproducible positioning in the mold. For other patients, the cast can interfere with some photon beam arrangements and placement of electron cones. For these reasons our methods are undergoing continual review. Coordinate transfer The Virtual Simulator provides a list of setup instructions for each beam. These instructions give couch position relative to the patient origin, and angle settings for gantry, couch, and collimator. To execute these instructions correctly, it is necessary to align the patient/cast with the couch, register the table position to the patient/ cast origin, and then execute the three couch translations. The patient/cast is automatically aligned with the couch by use ofjigs which attach to the couch. Our current prototypes (Fig. 6) are constructed from oak boards and include an adjustable footpiece. The jig is constructed to snugly accommodate the 2 ft wide Styrofoam substrate of our patient cast. Thus the jig provides automatic alignment of the cast to the x origin (table center) of the patient coordinate system and assures rotational alignment of the cast along all three axes. Jigs have been constructed for the CT table, the simulator, and all treatment machines. A jig with taller sides can be used as a frame during the casting process. In principle, the registration of the table position could be done by positioning the patient origin at isocenter and resetting mechanical or electronic table position readouts. This procedure would allow the setup instructions to be

1062

I. J. Radiation

Oncology

0 Biology 0 Physics

October

1990, Volume

19, Number

4

notch form

t

notches-

Fig. 5. (a) Polyurethane foam cast and form for creating notches in the cast that supports the head and neck of a patient. The notches are visible in the cast. (b) Without having to use attachment hardware, the notches lock a thermoplastic head cast to a foam cast. creating a single immobilization unit that surrounds the head but is easily removable in case of emergency.

execute d explicitly using the position readouts. In practice, we find that the table position readouts of our simulator and tre atment machines lack both the precision and accuracy necessary to be used as references for setting up patientr s “by the numbers.” We have found carefully aligned patient positioning lasers and rulers acceptable as referenc :es. Movable scales which attach to the positioning jig coul d also be used. * Alderson Phantom Patient, Alderson Research Laboratories. Inc., Stamford, CT 06904.

RESULTS The effectiveness of our rigidization and cot )rdil nate transfer methods was first tested using an an thrc ,pomorphic phantom* with articulated joints. Becau: se olFthe articulations, the phantom was scanned in a who Nle-blady cast. Selecting the bifurcation of the mainstem bl -0nc :hus as the tumor site, the Virtual Simulator was used to I3lan

Virtual

Fig. 6. Current coordinate of various lengths.

simulation

in the clinical setting 0 G. W. SHEROUSE eta/

transfer jig. Note movable

footpiece

that can be adjusted

1063

to accommodate

substrates

Fig. 7. A screen from the Virtual Simulator showing the setup of a small field on an anthropomorphic phantom with articulated joints. The right portion of the screen shows a beam’s eye view of the field, and the left shows two orthogonal views. Contours of the skin and bronchus are shown. The isocentric field required gantry, collimator, and table rotations as indicated by the “control panel” at the bottom center of the screen. Also, icons on the lower left show table and gantry positions as viewed from above and from the foot of the table.

1064

1. J. Radiation

Oncology

0 Biology 0 Physics

October

1990. Volume

19. Number

4

fields with isocenters at the bifurcation. In particular, one small, trapezoidal field was set using a combination of gantry rotation, collimator rotation, and table rotation (Fig. 7). In our routine clinical practice, this complex setup would be an unusual outcome of conventional simulation. This setup could have resulted only after a lengthy period of trial and error during conventional simulation, compared with a few minutes with the Virtual Simulator. After scanning, the phantom was removed from the cast to disturb the original scanning geometry and was repositioned by the technologist on the table of the conventional simulator. The phantom was set up by the technologist using directions printed out by the virtual simulation software. The instructions followed by the technologist were: (1) Adjust the phantom in the immobilization produce the scanning geometry.

cast to re-

(Step (2) places the origin of the patient coordinate at the simulator isocenter.)

system

(2) Adjust the table to align the reference marks immobilization cast with the laser beams.

on the

(Steps (3)-(5) are the lateral, vertical, and longitudinal table movements required to position the isocenter of the planned beam at the isocenter of the simulator. The table movements are the coordinates of the isocenter of the planned beam in the patient coordinate system.) (3) Looking toward the gantry, shift the table top .2 cm to the left. (4) Lower the table 11.9 cm. (5) Move the table top toward the gantry 22.3 cm. (6) Looking from above, rotate the table 36” counter clockwise. (7) Rotate the gantry 32” counter clockwise. (8) Looking into the collimator. rotate the collimator 39.4” counter clockwise. (9) Insert the tray with a lead-wire template representing the blocked field. 10) Take a radiograph. ( Figure 8 is the digitally reconstructed radiograph (DRR) beam, and Figure 9 is the actual Simulation film taken by the technologist on the first atempt. (The quality of the DRR in Fig. 8 is poor because of the low subject contrast of structures in the phantom. DRRs for real patients are of much higher quality (16).) Contours of the bifurcation. along with the field edges and crosshairs, are projected onto the DRR. On the conventional simulation film, the central ray of the beam clearly passes through the bifurcation, in agreement with the DRR. Using the tic marks on the crosshairs as reference points, positions of other landmarks. such as bolts, air cavities, and bones in the phantom, agree within 2 mm between the two radiographs. This test demonstrates that under ideal conditions, a treatment setup planned with the Virtual Simulator can

Fig. 8. Digitally reconstructed in Figure 7.

radiograph

(DRR) of field shown

be implemented with better than typical clinical precision using the immobilization and coordinate transfer methods discussed above. One implication is that virtual simulation may be able to replace conventional simulation,

CONCLUSIONS We have presented a practical system for clinical implementation of photon beam therapy planned with the Virtual Simulator or other similar systems for 3-dimensional treatment planning. The system depends on reproducible immobilization, which can be achieved with large polyurethane body casts and thermal plastic masks, and on a convenient jig arrangement for simplifying coordi-

( 16) showing the planned

Fig. 9. Conventional simulation film of field shown in Figure 7. The field was set up “by the numbers” as discussed in the text.

Virtual

simulation

in the clinical setting 0 G. W.

nate transfer. We have used an approach which emphasizes an appropriate balance between setup tolerances and impact on the clinical staff. In summary, some observations and conclusions can be drawn from our experience to date: ( 1) Beams designed by virtual simulation can be delivered to acceptable tolerances without the use of expensive materials and techniques. Virtual simulation has the potential to eliminate the trial and error of conventional simulation, and to reduce. or eliminate entirely, conventional simulation time. (2) Accurate planning and setup methods cannot compensate for poor immobilization techniques. Although the important issues are well understood and have been thoroughly discussed in the literature ( 12, 13. 18) there remains a considerable degree of art associated with

SHEROUSE

et al.

1065

the preparation of immobilization casts that function well. Immobilization is an open-ended problem that can always be improved, and the methods presented here are under constant scrutiny. (3) Careful design of the coordinate transfer jig is crucial. Our existing prototypes suffer from several drawbacks. The pin arrangements used for aligning the jig on the table are too fragile. A larger number of attachment points or a continuous attachment flange would be better. The walls of our jigs are too tall, causing stress on the jig, and sometimes patient discomfort, as patients maneuver into the cast: this can potentially interfere with certain beam directions. The 3” tall boards used for the frame also presented problems when scanning patients with target volumes below mid-abdomen. Movable coordinate scales for isocenter placement should be a (removable) part of the jig.

REFERENCES 1. Bauer-Kirpes, B.; Sturm. V.; Schlegel, W.; Lorenz. W. Computerized optimization of “‘1 implants in brain tumors. Int. J. Radiat. Oncol. Biol. Phys. 14: 1013-1023: 1988. 2. Chen, G. T. Y.; Goitein. M. Treatment planning for heavy charged particle beams. In: Advances in radiation therapy treatment planning, American Association of Physicists in Medicine Monograph No. 9. New York: American Institute of Physics: 1983: 5 14-54 1. 3. Cheng, C. W.: Chin. L. M.: Kijewski, P. K. A coordinate transfer of anatomical information from CT to treatment simulation. Int. J. Radiat. Oncol. Biol. Phys. 13: 1559-1569; 1987. 4. Findlay, P. A.: Wright, D. C.; Rosenow. U.: Harrington, F. S.: Miller. R. W. “‘1 interstitial brachytherapy for primary malignant brain tumors: technical aspects of treatment planning and implantation methods. Int. J. Radiat. Oncol. Biol. Phys. 11:2021-2026; 1985. 5. Fraass, B. A.: McShan, D. L. 3-D treatment planning: I. Overview of a clinical planning system. In: The use of computers in radiation therapy, Proceedings of the Ninth International Conference, Scheveningen. The Netherlands: North Holland Publishing Co.: 1987:273-276. 6. Goitein. M.; Abrams, M. Multidimensional treatment planning: I. Delineation of anatomy. Int. J. Radiat. Oncol. BioI. Phys. 9:777-787; 1983. Goitein, M.: Abrams, M.; Rowell, D.; Pollari. H.: Wiles, J. Multidimensional treatment planning: II. Beam’s eye-view. back projection, and projection through CT sections. Int. J. Radiat. Oncol. Biol. Phys. 9:789-797: 1983. Hartmann. G. H.; SchIegeI, W.; Sturm, V.; Lorenz. W. J. A fast algorithm to calculate three dimensional dose distributions in radiosurgery. In: The use of computers in radiation therapy, Proceedings of the Eighth International Conference. Toronto, Canada. Silver Spring, MD: IEEE Computer Society Press: 1984:99- 102. Lutz, W.: Winston, K. R.: Maleki. N. A system for stereotactic radiosurgery with a linear accelerator. Int. J. Radiat. Oncol. BioI. Phys. 14:373-381; 1988.

10 Mohan, R.: Barest, G.: Brewster, L.: Chui, C.: Kutcher. G.: Laughlin. J.; Fuks. Z. A comprehensive three-dimensional radiation treatment planning system. Int. J. Radiat. Oncol. Biol. Phys. 15:481-495; 1988. 11. Mosher, C. E.; Sherouse, G. W.: Mills. P. H.; Novins. K. L.; Pizer, S. M.; Rosenman, J. G.: Chaney, E. L. The virtual simulator. In: Proceedings of the I986 workshop on interactive computer graphics. New York: Assoc. for Comp. Machinery (ACM): 1987:37-42. 12. Rabinowitz. I.; Broomberg, J.: Goitein, M.; McCarthy, K.; Leong, J. Accuracy of radiation field alignment in clinical practice. Int. J. Radiat. Oncol. Biol. Phys. 11:1857-1867; 1985. 13. Reinstein, E. R.; Meek, A. G. Workshop on geometric accuracy and reproducibility in radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 13:809-810: 1987. 14. Rosenman, J. R.: Sherouse, G. W.: Chaney. E. L.; Tepper, J. E. Virtual simulation: initial clinical results. Int. J. Radiat. Oncol. Biol. Phys. (In press). 15. Sherouse. G. W.; Mosher. C. E.; Novins. K.: Rosenman, J.; Chaney. E. L. Virtual simulation: concept and implementation. In: The use of computers in radiation therapy, Proceedings of the Ninth International Conference, Scheveningen. The Netherlands: North Holland Publishing Co.: 1987:433-436. 16. Sherouse. G. W.; Novins. K.; Chaney, E. L. Computation of digitally reconstructed radiographs for use in radiotherapy treatment design. Int. J. Radiat. Oncol. Biol. Phys. I8:65 I658; 1990. 17. Ten Haken. R. K.; Diaz, R. F.; McShan. D. L.; Fraass, B. A.; Taren, J. A.; Hood, T. W. From manual to 3-D computerized treatment planning for “‘1 stereotactic brain implants. Int. J. Radiat. Oncol. Biol. Phys. 15:467-480: 1988. 18. Verhey, L. V.: Goitein. M.: McNulty. P.: Munzenrider, J. E.: Suit, H. D. Precise positioning of patients for radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 8:289-294: 1982.

Virtual simulation in the clinical setting: some practical considerations.

Virtual simulation departs from normal practice by replacing conventional treatment simulation with 3-dimensional image data and computer software. Im...
1MB Sizes 0 Downloads 0 Views