Inr J Radrarm Oncology Bml Phu Vol. 20. PP. 843-85 Pnnted in the Ll S.A. All rights reserved.

0360-3016191

I

t .oO

$3.00 Press plc

Copyright 0 I99 I Pergamon

??Technical Innovations and Notes

VIRTUAL JULIAN ROSENMAN,

SIMULATION:

PH.D.,

M.D.,*

INITIAL

CLINICAL

RESULTS

SCOTT L. SAILER M.D.,+ GEORGE W. SHEROUSE, M.S.,+

EDWARD L. CHANEY,

PH.D.+ AND JOEL E. TEPPER, M.D.+

University of North Carolina at Chapel Hill We have developed a graphics-based three-dimensional treatment design system that permits the physician to easily understand which anatomy will be treated for any arbitrary beam orientation. Our implementation of this system differs from others in that the software (the Virtual Simulator) simulates the full functionality of a (physical) radiation therapy simulator allowing it to be easily used by physicians. The details of the of our initial clinical experience with virtual simulation are presented in this paper. Virtual simulation was attempted in 71 patients and completed in 65. In 41/71 patients (58%), the beam orientations chosen differed significantly from those traditionally used in our department. Although virtual simulation lead to traditional radiation portals in the remaining patients, in 23/71 (32%) secondary blocking was designed which was different from that which would have been conventionally employed. Thus, overall, virtual simulation lead to treatment changes in 64/71 (90%) of the patients in whom it was attempted. In 78% of evaluable patients the treatment designed with virtual simulation could be implemented on the physical simulator with a precision of +5 mm (+3 mm for brain and head and neck). Thus virtual simulation allowed both accurate planning and execution of treatment plans that would be difficult to achieve with conventional methods. Treatment planning, Computer graphics, 3D displays.

reports, errors in tumor targeting were found in 195/402 (49%) patients. These results were determined by comparing the actual treatment setups with data obtainable by computed tomography (CT). Other studies have confirmed this pessimistic observation (13). Webb et al. (40) compared single plane to multiplane planning for treatment of the intact breast, and showed that the dose distribution throughout the breast that was commonly achieved using standard planning techniques often conformed poorly to the underlying anatomy. Although it seems straightforward that inaccurate tumor targeting should lead to a poor clinical outcome, there is surprisingly little data to support this claim. However, it has been demonstrated that lack of treatment accuracy can have an adverse effect on patients with cancer of the nasopharynx ( 14), lung (5,2 1,40), and Hodgkin’s Disease (12). Goitein (8) has estimated that the elimination of geometric misses in a general radiation therapy population might increase cure rates by 3-4%. At present, CT, magnetic resonance imaging (MRI), and other imaging data are often used to augment the information present on simulation films. The approximate

INTRODUCTION

of the radiation oncologist to control the localregional component of a malignancy reduces or eliminates the possibility of cure, and often results in a substantial decrease in the quality of a patient’s life. It has been estimated that if local tumor control were always achieved, more than 40,000 lives would be saved annually in the U.S. (2, 3 1, 32). Failure to achieve local-regional tumor control after irradiation can occur because of biologic or technical factors. Some tumors (such as glioblastoma) can not be controlled with any radiation dose that we are likely to achieve; future progress in their treatment will require advances in radiation response modifiers, altered fractionation schemes, the use of high LET radiation, or other strategies. There is evidence, however, that geometric tumor misses (inaccurate tumor targeting), or the more subtle failure to match the dose distribution with the target volume satisfactorily (poor conformation of dose distribution and tumor) may be fairly common. For example, in a review by Tepper and Padikal (38) of six published Failure

Presentedat the ASTRO Annual Meeting,New Orleans, 1988.

This work was sponsored, in part, by grants from the Whitaker Foundation and the Siemens Corporation. Accepted for publication 12 October 1990.

* Departments of Radiation Oncology and Computer Science. + Department of Radiation Oncology. Reprint requests to: Dr. Julian Rosenman, Department of Radiation Oncology, University of North Carolina Hospitals, Chapel Hill, NC 275 16. 843

x44

I.

J.

Radiation

Oncology 0

Biology0

Physics

spatial position of tumor volumes and radiosensitive organs, as inferred from these scans. are drawn on simulator films by the physician. However, accurate registration of the CT data with the simulator film. while allowing for differences in patient positioning between simulation and CT scanning and for beam divergence, cannot be consistently and accurately achieved without the appropriate computer software and hardware graphics tools. We report here our early clinical experience with such a set of tools in use at the University of North Carolina.

METHODS

AND

MATERIALS

A technical description of the virtual simulation process has been given in the nonmedical literature (28). but a brief review will be given here so as to allow clinicians to understand the general approach. Following is an outline of the key steps in our treatment planning process: I. The first step is to gather imaging data of the patient in such a way as to permit precise transfer of the setup coordinates from the planning center to the treatment room. Our approach is to CT scan the patient on a flat table top while he/she is immobilized in the treatment position. A reference mark is placed on the cast at the level of the first CT slice for later use in treatment setups. We use a scanner* with a 70 cm gantry opening and a 52 cm scan circle. 2. The CT scans are then transferred to the planning workstation to be prepared for processing and display. Objects of significance to the planning process such as dose critical organs, tumor volumes, or bones are then contoured on the CT scans. Abstraction of these essential structures from 3D image sets would require many physician-hours to accomplish if automated methods were not available. We have recently implemented an image contouring program under the X Window System (26) that allows for an interactive combination of manual and automatic contouring ( 17). 3. The radiation portals are then designed using the “Virtual Simulator,” software that fully implements the function of a physical simulator. The Virtual Simulator operates in a similar manner to the physical simulator but, in addition, gives the user the ability to explore a wide variety of treatment geometries rapidly. As descrip-

.April 109 I. Volume 3). Number 4

tions of the Virtual Simulator have been published only in the non-medical literature ( 19, 27. 28) a brief overview is provided here: The user of the Virtual Simulator (most often the physician) is provided with a 3D display of the patient model which consists of the tumor volume and relevant anatomical structures abstracted from the CT data. Views of the simulator gantry and table from three orthogonal views are also provided. These views can be resized and moved anywhere on the screen as desired. Figure I shows the Virtual Simulator screen as it appears for the design of a brain tumor portal. The beam is shaped on the model of the patient’s head (right upper window), which includes the skin (blue) and the eyes (green). Radiation portal design on the Virtual Simulator usually begins by placing the target volume at isocenter by moving the patient model along any of the three orthogonal axes. This process is actually easier to do on the Virtual Simulator, where the target volume is visible. than on the physical simulator, where it usually is not. Beams can then be arranged at arbitrary angles (in 3 dimensions) around the target with full knowledge of the impact of each beam on the defined normal structures. Since all coordinate system transformations and rotation angles are handled by the software. the user is free to concentrate on the more relevant issues of proper tumor targeting. The held itself is then shaped using a mouse or other pointing device (Fig. 3): in this case a conedown ofa brain tumor portal is being designed. When the beam outline is closed. the user can view the beam overlaid on any or all of the original CT slices (lower left corner). Once a radiation beam is satisfactorily placed, provisions are available to copy and oppose it automatically, or create other beams. When all beams have been designed, the virtual simulation is complete and, like its physical counterpart. the Virtual Simulator can create output. 4. Digitally reconstructed radiographs are then produced for each field of the final plan. Projections of the tumor. target volume, and other structures of interest are automatically transferred from contours on the CT scans to the digitally reconstructed radiographs. 5. The Virtual Simulator then produces a list of beam and table parameters needed to reproduce each specified portal. All of these numbers are absolute (for example.

Fig. 1. A typical Virtual Simulator screen. An oblique brain tumor portal is being designed on the patient model (upper right). Shown are contours of the skin (blue), eyes (green). and tumor (red). In the lower left corner. the tumor target is overlaid on a CT slice. In the upper left corner are two orthogonal views of the patient models. Any of these “windows” can be resized or moved as desired. Not shown are hidden windows that allow the Virtual Simulator “gantry”, “collimater” or “table” to be moved. Fig. 2. The brain tumor left hand corner.

* Siemens

DR/G.

portal being completed.

The radiation

beam is now projected

on the CT slice in the lower

846

I. J. Radiation

Oncology

0 Biology ??Physics

Table 1. Patients

April 199 I, Volume 20, Number 4

planned

with virtual simulation Changes

Fields planned

Site

Number

Whole treatment

with the virtual simulator Boost field

Plan not used

Beam position

in plan due to virtual simulation Block position

No change

Abdomen Brain Extremity Head and neck Pelvis Spine Thorax

7 19 2 7 3 1 4

1

0

21 2 15 9 2 14

2 0 7 4 0 8

0 0 1 2 1 2

6 15 2 6 1 1 10

2 5 0 8 6 0 2

0 1 0 0 0 0 0

Total

71

43

22

6

41

23

I

8

gantry angle) except for longitudinal table displacement, which is relative to the original reference mark. Other output of the virtual simulator includes construction specifications and templates for beam modifiers such as blocks and compensating filters. 6. Following virtual simulation and the production of the digitally reconstructed radiographs, the patient is put into the body cast and placed on the physical simulator. The fields are set up using only the instructions generated during virtual simulation, a process we call “setting up by the numbers.” The physical simulation films are then compared to the digitally reconstructed ones as a quality control. For sites in the head and neck, a displacement of the field isocenter or edge of 3 mm or less between the two was considered acceptable (23); for other, less easily mobilized sites, a 5 mm tolerance was considered acceptable. Data for this study were collected in part prospectively (from a form that is filled out after the virtual simulation is completed) and part retrospectively (by chart review). RESULTS

Overview Seventy-one patients have undergone a complete virtual simulation between December 1987 and November 1989. Table 1 provides an overview of these patients by anatomic sites. Under the column labeled “Fields planned with the Virtual Simulator,” patients were subdivided into those whose entire treatment was planned on the Virtual Simulator and those who only had their boost field so planned. In six patients the virtual simulation plan could not be completed or implemented. In four of these patients there were technical or immobilization problems, and two patients could not hold still long enough to be scanned in the treatment position. The column labeled “Changes in plan due to virtual simulation” records the attending clinician’s best guess as to how the patient’s treatment differed from what it would have been had only conventional simulation been avail-

able. In 41/7 1 (58%) of the patients, virtual simulation resulted in radiation beam orientation(s) that differed from those traditionally used in our department. In 23/7 1 (32%) of the patients, standard radiation portals were used, but it was felt that the shielding blocks designed with the aid of the Virtual Simulator were different and more accurate than they would have been if they had been designed by traditional means. In one patient with a pituitary adenoma, virtual simulation resulted only in traditional beam orientations and blocks, and in six patients the virtual simulation plans could not be implemented for various technical reasons. However, in 64/7 1 (90%) of the patients, virtual simulation lead to substantial changes in the treatment plan. Thirty-two patients had prospective data available on the precision of the clinical set-ups after virtual simulation (Table 2). For 25/32 (78%) of the patients, we were able to transfer the virtually designed radiation portals to the patient with a clinical precision of 15 mm (13 mm for brain and head and neck tumors) and to verify it with a simulation check film. Figures 3 and 4 show, respectively, the digitally reconstructed radiograph, and (physical) simulator film of the brain tumor portal being designed in Figures 1 and 2. In 6/32 (19%) cases, the replication precision was not acceptable; the cause was usually traced to patient movement within the cast or an insufficiently rigid cast.

Brain tumors From a technical perspective, virtual simulation was easy to use for brain tumors because only a few anatomic structures such as tumor, surrounding edema, skull, eyes, optic nerves, and sometimes spinal cord needed to be abstracted and manipulated. Usually 40-60 CT slices were taken at 4 mm intervals. In 15/2 1 brain tumor cases, an alternative setup using more than two fields was produced using virtual simulation. In addition to the standard lateral opposed ports, a third (or even fourth) field from an anterior and/or vertex approach was often used. These setups resulted in a radiation dose distribution which had better tumor confor-

Virtual simulation

Table 2. Precision in transferring virtually designed radiation oortals to the oatients Precision Site

Number of evaluable cases

Abdomen Brain Extremity Head and neck* Pelvis Spine Thorax

3 10 2 8 5 1 3

2 9 0 7 3 1 3

1 1 2 1 2 0 0

Total

32

25

7

5 mm

* 53 mm for brain, and head and neck sites.

mation than would have been possible using opposed lateral fields. In particular, the ability to custom block the vertex field was found to spare irradiation of lateral normal brain. Such blocking would be nearly impossible using conventional treatment planning techniques. In five other patients the tumors were so large that opposed lateral fields appeared to be optimal. However, we felt that the custom blocks were more reliably designed using virtual simula-

0 J. ROSENMAN etal.

847

tion than would have been accomplished with conventional simulation. Technical problems were encountered in two of the successful virtual simulations: in one case the physical simulation film did not match the digitally reconstructed one because of a misinterpretation of the setup instructions. In the other case the patient was initially set up with left and right reversed, which underscored the obvious need for the clinician to look at the final setup on the patient.

Head and neck tumors Head and neck tumors reside in an area of complex anatomy, and should, therefore, be good candidates for virtual simulation. A case in point is that of an elderly female with a T3N0 squamous cell carcinoma of the right vocal cord. At simulation it was found that because of her extremely short neck, almost all of the lateral fields would have to pass through her shoulders. Furthermore, osteoarthritis made it impossible for her to extend her head far enough for the tumor to be treated AP-PA. After experimentation it was found that positioning the gantry 10” above the true lateral position, with the foot of the treatment table “kicked out” 15” away from the treatment

Fig. 3. The digitally reconstructed radiograph of the brain tumor portal designed tumor target volume, central axes, and portal outlines are automatically overlaid.

in Figures

1 and 2. The brain

848

I. .I. Radiation Oncology 0 Biology 0 Phq’sics

April

I YY I. Volume

20. Number

4

To date, the greatest advantage of virtual simulation in head and neck cancer patients appears to be for block placement. In 8/ 15 cases the clinician felt that the virtually designed blocks were different than those which would have resulted from conventional treatment planning. However, in difficult areas to treat, such as the paranasal sinuses, nasopharynx, and the middle ear, virtual simulation permitted innovative alternative plans.

Fig. 4. The simulation radiograph. The patient was set up using the directions output from the Virtual Simulator. The digitally reconstructed and physical simulation radiograph match each other within 5 mm. beam, would allow the light field to shine on all the areas of the neck that needed to be treated. Simulation films were taken, but it was impossible to determine the site of her larynx because of the unfamiliar confluence of shadows seen at this compound angle. In addition, there was some uncertainly about the position of the spinal cord. The patient was treated with large fields that covered both necks down to the clavicle to 4500 cGy. For the boost field the patient underwent virtual simulation. At that time it was found that the larynx projected more inferiorly on the simulation film than had been thought. Although the tumor was adequately covered with the large field, it would have been missed on the (pre-virtual simulation) proposed boost field. The patient was treated with the virtually simulated boost field to 6800 cGy and is without evidence of disease 15 months after completing therapy. Another patient whom we felt substantially benefited from virtual simulation was an elderly male with a far advanced middle ear carcinoma invading the temporal bone and extending to the posterior ethmoids. It was elected to treat him with a wedge pair with secondary blocking to reduce the amount of brain treated. Designing shaping blocks for a wedge pair treatment, a notoriously difficult problem, ( 16) was straightforward with the Virtual Simulator. Because most of the temporal lobe could be spared, it was elected to treat the patient to 6800 cGy. He is without evidence of disease at 4 months.

Other patients thought to have profited from the virtual simulation included five patients with prostate cancer, four of whom had their boost held(s) designed with the aid of virtual simulation. Conventional (4-field) setups were used. but in one patient the gland was abnormally shaped, with very posterior extension of the seminal vesicles. The spatial position of the prostate was not appreciated until after production of the digitally reconstructed radiographs. and likely would have been inaccurately targeted had the virtual simulation not been done. In many of the lung cancer patients it was found that the off-cord boosts were far easier to design with the Virtual Simulator than with conventional simulation. After defining on CT the amount of mediastinum to be boosted. it was straightforward to determine the optimal angle for a pair of opposed obliqued beams for sparing both the spinal cord and as much contralateral lung as possible. For patients with large abdominal fields, it was often found that lateral fields could be angled so as to spare a substantial amount of one or both kidneys.

DISCUSSION The concept of tightly integrating CT scans into the treatment planning process is not new (24) and such systems are currently under development at a number of institutions (1. 4, 9-l I, 15, 18, 20, 30). However, our system differs from these in two important ways: the manmachine interface and portability of the software. We have paid a great deal of attention to the user interface for our system: the goal is to have the clinician operate the entire system with only modest intervention of the physicists. This is not a minor point of convenience; we feel that the physician who has examined and studied the patient is most capable of making the many small decisions that heavily influence the treatment planning process. To make our virtual simulation process comfortable and familiar to the physician. and thus to provide maximum support for the decision-making process, we have chosen to make virtual simulation as much like physical simulation as possible. Thus, the physician already knows how to operate the Virtual Simulator and can concentrate on the clinical issues. This use of software as a metaphor for a familiar system is not a new idea: Apple Computer Inc. has based the successful Macintosh’” computer on a desk top metaphor, for example. However, to our knowledge, this approach is new to radiation treat-

Virtual simulation

ment planning. The success of this approach is measured by the fact that all of the attending physicians, and most of the residents, are now able to operate the system with little or no help. Because computer hardware capable of running our projected versions of the Virtual Simulator (higher quality images, faster response time, and other features) does not yet exist, we choose to adhere to software standards that will allow our applications to migrate to new hardware as it becomes available (3). In most institutions, treatment planning software is highly machine-specific. Our software is written in a modular form in the C programming language, and runs under the X Window System (26) in a UNIX’” environment. It has been run on a wide range of computers (most computers support UNIX and the X Window System) with little or no alteration in the computer code. The Virtual Simulator and additional support code has been licensed by more than 25 other institutions and is available to other radiation therapy departments.*

Because 3D treatment planning is expensive and technically demanding, it must be demonstrated that these techniques can lead to better treatment outcome for some subgroups of patients. Before any realistic clinical trials can begin. however, these subgroups must be identified. However, only a few reports detailing how 3D planning systems can alter clinical practice have yet appeared (6, 29, 33-37, 39). Furthermore, these reports are mostly abstracts (6, 29, 34. 35, 37, 39)-many are not reports of actual clinical usage-and represent the experience of only a few institutions. Therefore, we feel it important to present the results of our early clinical usage of 3D treatment planning so as to help define the utility of this techniques in clinical practice. Our initial experience has suggested that the Virtual Simulator allows the physician to improve the quality of radiation therapy treatment planning in several ways. First, for some patients, virtual simulation permits more tumor targeting accuracy than is possible with the best conventional techniques. An example is the prostate cancer patient with very posterior seminal vesicles that probably would have been missed with a conventional simulation. Second, the Virtual Simulator allows the physician to design alternative beam arrangements that can represent an improvement over conventional treatment plans. Perhaps the most successful example of the value of an alternative beam arrangements was seen in the patient with an advanced laryngeal cancer. It is unlikely that she could have been cured with a conventional treatment plan. As shown in Table 1, 4 l/7 1 virtually simulated patients were treated with non-standard beam arrangements. Although we cannot categorically state that these alternative treatments produced higher cure rates, they were felt by

* Please contact G. W. Sherouse, care of this department.

0 J ROSENMAN (‘I ul.

849

the treating physician to be superior to conventional plans. Finally, an obvious advantage of virtual simulation is that it permits the physician to draw customized secondary blocking with confidence, even when the radiation beams are entering at unusual angles. Examples include the shaping of vertex brain and wedge pair portals. In these situations, secondary blocking would ordinarily have been used sparingly, if at all, for fear of blocking tumor. We had initially been concerned about our ability to carry out the setups defined by the Virtual Simulator. However, as shown in Table 2, in 25/32 (78%) ofthe cases the physical simulation radiograph matched the digitally reconstructed one within clinically acceptable tolerances. In most cases where the setup precision exceeded clinical tolerance, the problem was found to be caused by human error. Setup problems have become progressively less frequent as our technologists have better understood our system and as minor technical improvements were made. The relaxation of traditional constraints on radiation beam placement made possible by the virtual simulation process should allow for a closer “fit” between radiation dose distribution and tumor volume (better dose conformation). which in turn could result in either lower treatment morbidity or higher tumor doses for a given rate of patient morbidity.

Perhaps the most vexing problem to be solved in 3D treatment planning is that of automatic data abstraction from CT scans. Currently this process takes l-2 hr for each patient despite our use of automatic edge tracking software. Pizer and co-workers (22) have been making impressive progress in solving this problem by using multiscale global/local, methods that model human vision rather than purely local object recognition methods. These techniques could ultimately prove valuable to 3D treatment planning. We feel that it is necessary for the user to be able to manipulate a high quality screen image of the patient model smoothly without undue delays. Unfortunately, speed of update and image quality are diametrically opposed goals. In its current implementation, the Virtual Simulator uses the relatively low quality wire-loop display with rapid display time. We ultimately plan to implement surface shaded displays or volume rendering (7, 25) to make the displays more realistic. SztmmarJ Virtual simulation has proved to an effective and clinically practical method of designing radiation portals directly from CT data. The main strength of this method of implementation is that it is already familiar to the physician. Designing portals on the Virtual Simulator is much like designing them on the physical simulator. The output

X50

I.

J.

Radiation

Oncology 0 Biology 0 Physics

of the Virtual Simulator consists of digitally reconstructed radiographs and treatment machine settings which are obvious analogues of simulation films and handwritten

.April 199

I,

Volume

20. Number

4

setup instructions. Virtual simulation, therefore. is not so much a new approach to treatment planning as an extension and improvement of conventional techniques.

REFERENCES I. Chen. G. T. Y.: Goitein, M. Treatment planning for heavy charged particle beams. Advances in radiation therapy treatment planning, AAPM Monogram #9. NY: American Institute of Physics, 1983:5 14-541, 2. Coleman, C. N. Hypoxic radiosensitizers: expectations and progress in drug development. Int. J. Radiat. Oncol. Biol. Phys. I 1:323-329: 1985. 3. Corcoran. E. Picture computation: computer makers vie to build “graphics supercomputers.” Sci. American 259:9798; August, 1988. 4. Dahlin, H.; Lamm. I. L.: Landberg. T.: Levernes, S.; Ulso. N. User requirements on CT-based computed dose planning systems in radiation therapy: Presentation of “check lists.” Comp. Prog. Biomed. 16:131-138: 1983. 5. Doss, L. L. Localization error and local recurrence in upper airway carcinoma. In: Reinstein, L. E. eds. Proceedings of the Workshop on Quality Control in Radiotherapy Departments. CALGB: 1979. 6. Emami. B.; Purdy. J. A.: Manolis. J.: Barest, G.: Cheng, E.: Coia, L.: Doppke. K.: Calvin. J.: Lo Sasso, T.: Matthews J.; Munzenrider, J.; Shank, B. Three dimensional treatment planning for lung cancer (Abstr.). Int. .I. Radiat. Oncol. Biol. Phys. I S(Suppl.): 147: 1988. 7. Fuchs, H.: Pizer, S. M.: Creasy, J. L.; Renner. J. B.: Rosenman, J. G. Interactive. richly cued display of multiple 3D objects in medical images. Proc. SPIE Conf.. Med. Imag. II 914:842-849; 1988. 8. Goitein. M. The utility of computed tomography in radiation therapy: an estimate of outcome. Int. J. Radiat. Oncol. Biol. Phys. 5: 1799- 1807; 1979. 9. Goitein. M.; Abrams, M. Multi-dimensional treatment planning: I. Delineation of anatomy. Int. J. Radiat. Oncol. Biol. Phys. 9:777-787; 1983. IO. Goitein, M.; Abrams, M.: Rowell. D.: Pollari, H.: Wiles. J. Multi-dimensional treatment planning: II. Beam’s eye-view. back projection. and projection through CT sections. Int. J. Radiat. Oncol. Biol. Phys. 9:789-797: 1983. 1I. Houlard. J. P.; Dutreix. A. 3D display of radiotherapy treatment plans. In: Proceedings of the Eighth International Conference on the Use of Computers in Radiation Therapy. Silver Spring, MD: IEEE Computer Society Press: 1984: 219. 12. Kinzie, J. J.; Hanks, G. E.: Maclean. C. J.: Kramer. S. Patterns of care study: Hodgkin’s disease relapse rates and adequacy of portals. Cancer 52:2223-2226: 1983. 13. Lichter. A. S.; Fraass, B. A.: van de Geijn. J.: Fredrickson, H. A.; Glatstein, E. An overview of clinical requirements and clinical utility of computed tomography based radiotherapy treatment planning. In: Ling. C. C., Rogers. C. C.. Morton, R. J. eds. Computed tomography in radiation therapy New York: Raven Press: 1983: l-2 I. 14. Marks, J. E.: Bedwinek, J. M.: Lee, F.: Purdy. J. A.; Perez, C. A. Dose-response analysis for nasopharyngeal carcinoma. Cancer 50: IO42- 1050; 1982. 15. McShan. D. L.; Fraass. B. A.: Lichter, A. S. Three dimensional portal design and verification (Abstr.). Med. Phys. 13:575: 1986. 16. Million, R. R.. Cassisi, N. J. Major salivary gland tumors. In: Million, R. R., Cassisi. N. J.. eds. Management of head and neck cancer: a multidisciplinary approach, Philadelphia: J. B. Lippincott Co.: 1984:541.

17. Mills, P. H.: Fuchs, H.: Pizer, S. M.; Rosenman, J. G. IMEX: A tool for image display and contour management in a windowing environment. Med. Imag. III: image capture and display. Proc. SPIE 1091:132-142; 1989. 18. Mohan, R.; Barest, G.; Brewster. L. J.; Chui, C. S.; Kutcher. G. J.; Laughlin. J. S.; Fuks. Z. A comprehensive three-dimensional radiation treatment planning system. Int. J. Radiat. Oncol. Biol. Phys. l5:48 l-495: 1988. 19. Mosher, C. E.; Sherouse. G. W.; Chaney, E. L.; Rosenman. J. G. 3D displays and user interface design for a radiation therapy treatment planning CAD tool. Proc. SPIE. threedimensional imaging and remote sensing imaging. Los Angeles, CA 902:64-72; 1988. 20. Nagata, Y.; Nishidai. T.: Abe, M.: Takahashi, M.; Yukawa, Y.: Nohara, H.; Yamaoka, N.; Saida. T.; Ishihara. H.; Kubo. Y.: Ohta. H.: Inamura. K. CT simulator: A new treatment planning system for radiotherapy (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. 13(Suppl.): 176: 1987. 2 I. Perez. C. A.; Stanley, K.; Grundy, G.; Hanson. W.; Rubin. P.: Kramer, S.: Brady. L.; Marks, J.; Tamayo, R. P.; Brown, G. S.: Concannon, J. P.: Rotman. M. Impact of radiation technique and tumor extent in tumor control and survival of patients with unresectable non-oat cell carcinoma of the lung: report by the RTOG. Cancer 50: 109 I-1099: 1982. 22. Pirer. S. M.: Gauch. J. M.: Lifshitz. L. M.; Oliver, W. Interactive 2-D and 3-D object defmition in medical images based on multiresolution image descriptions. SPIE Proc. 9 14:438-449: 1988. 23. Rabinowitz, I.: Broomberg. J.: Goitein. M.; McCarthy, K.; Leong. J. Accuracy of radiation field alignment in clinical practice. Int. J. Radiat. Oncol. Biol. Phys. I I: 1857-1867; 1985. 24. Reinstein, L. E.; McShan, D.: Webber, B. M.; Glicksman, A. S. A computer assisted three-dimensional treatment planning system. Radiology I27:259-264: 1978. 25. Rosenman, J.: Sherouse. G. W.; Fuchs. H.: Pizer. S. M.: Skinner, A. L.: Mosher. C.: Novins, K.; Tepper. J. E. Threedimensional display techniques in radiation therapy treatment planning. Int. J. Radiat. Oncol. Biol. Phys. 16:263269; 1989. 26. Scheifler, R. W.; Gettys. J. The X window system. ACM Trans. Graph. 5:79- 109; 1986. 27. Sherouse, G. W.; Mosher, G. User interface issues in radiotherapy CAD software. In: Bruinvis. I. A. D, van der Giessen, P. H., van Kleffens, H. J., Wittkamper. F. W., eds. Proceedings of the 9th International Conference on the Use of Computers in Radiation Therapy. Amsterdam, The Netherlands: Elsevier Science Publishers B.V.; 1987:429-432. 28. Sherouse. G. W.; Mosher. C. E.; Novins, K.; Rosenman, J.; Chaney, E. L. Virtual simulation: Concept and implementation. In: Bruinvis. I. A. D.. van der Giessen, P. H., van Kleffens. H. J.. Wittkamper. F. W.. eds. Proceedings of the 9th International Conference on the Use of Computers in Radiation Therapy. Amsterdam, The Netherlands: Elsevier Science Publishers B.V.; 1987:433-436. 29. Simpson, J. R.; Purdy, J. A.; Manolis, J. M.; Burman. C.; Forman. J.; Fuks, Z.; Cheng, E.; Chu, J.: Matthews, J.; Mohan, R.: Pilepich, M. V.: Solin, L.; Tepper, J.; Urie. M. Three dimensional treatment planing considerations for prostate cancer (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. I S(Suppl.): 147: 1988.

Virtual simulation 0 J. ROSENMANeful. 30. Smith. R. M.; Calvin, J. M.; Harris, R. Correlation of CT information with simulation and portal radiographs (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. 13(Suppl.): 176; 1987.

37.

31. Suit, H. D.; Westgate, S. J. Impact of improved local control on survival. Int. J. Radiat. Oncol. Biol. Phys. 12:453-458; 1986. 32. Suit. H. D.; Tepper, J. E. Impact of improved local control on survival in patients with soft tissue sarcoma. Int. J. Radiat. Oncol. Biol. Phys. 12:699-700: 1986. therapy to improve the 33. Tate, T.; Shentall, G. Conformation irradiation of the spinal axis. Int. J. Radiat. Oncol. Biol. Phys. 16:505-5 10: 1989. 34. Ten Haken, R. K.: McShan, D. L.; Fraass. B. A.: Archer, P. G.: Lichter. A. S. Practical use of 3-D treatment planning for non-coplanar beams in the abdomen (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. 1S(Suppl.): 144: 1988. 35. Ten Haken, R. K.; Forman, J. D.; Gerhardsson. A.: Heimberger, D. K.; McShan. D. L.: Perrez-Tamayo. C.; Schoeppel, S. L.; Tesser, R. J.; Lichter. A. S. Movement of the prostate in response to the state of filling of the rectum and bladder (Abstr.). Int. J. Radiat. Oncol. BioI. Phys. 17:123: 1989. 36. Ten Haken. R. K.: Perez-Tamayo, C.: Tesser. R. J.: McShan, D. L.; Fraass, B. A.; Lichter. A. S. Boost treatment of the

38.

39.

40.

41

851

prostate using shaped fixed fields. Int. J. Radiat. Oncol. Biol. Phys. 16: 193-200; 1989. Tesser, R. J.; Carrell, M. B.; Ten Haken, R. K.; Laurence, T. S. The application of 3D treatment planning and dose volume histograms in the treatment of intrahepatic malignancies (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. 17:127: 1989. Tepper, J. E.; Padikal, T. N. The role of computed tomography in treatment planning. In: Bleehen, N. M., Glastein, E., Haybittle, J. L., eds. Radiation therapy planning. New York, NY: Marcel Dekker, Inc: 1983: 139- 158. Thornton, A. F.; Hegarty, T. J.; Archer, P.: McShan, D. L.: Ten Haken, R. K.; Lichter, A. S. Three-dimensional treatment planning of astrocytomas-a dosimetric study of cerebral radiation (Abstr.). Int. J. Radiat. Oncol. Biol. Phys. 17:125: 1989. Webb, S.; Leach, M. 0.; Bently, R. E.; Maureemootoo, K.; Yarnold, J. R.; Toms, M. A.; Gardiner, J.; Parton. D. Clinical dosimetry for radiotherapy to the breast based on imaging with the prototype Royal Marsden Hospital CT simulator. Phys. Med. Biol. 32:835-845: 1987. White, J. E.: Chen, T.; McCracken, J.: Kennedy, P.: Seydel. H. G.; Hartman. G.: Mira, J.; Khan, M.; Durrance, F. Y.; Skinner, 0. The influence of radiation therapy quality control on survival, response, and sites of relapse in oat cell carcinoma of the lung. Cancer 50:1084-1090; 1982.