- Email: [email protected]
Methodologies and tools for proton beam design for lung tumors
Int. J. Radiation Oncology Biol. Phys., Vol. 49, No. 5, pp. 1429 –1438, 2001 Copyright © 2001 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/01/$–see front matter
PII S0360-3016(00)01555-8
PHYSICS CONTRIBUTION
METHODOLOGIES AND TOOLS FOR PROTON BEAM DESIGN FOR LUNG TUMORS MICHAEL F. MOYERS, PH.D., DANIEL W. MILLER, PH.D., DAVID A. BUSH, M.D., JERRY D. SLATER, M.D.
AND
Department of Radiation Medicine, Loma Linda University Medical Center, Loma Linda, CA Purpose: Proton beams can potentially increase the dose delivered to lung tumors without increasing the dose to critical normal tissues because protons can be stopped before encountering the normal tissues. This potential can only be realized if tissue motion and planning uncertainties are correctly included during planning. This study evaluated several planning strategies to determine which method best provides adequate tumor coverage, minimal normal tissue irradiation, and simplicity of use. Methods and Materials: Proton beam treatment plans were generated using one or more of three different planning strategies. These strategies included designing apertures and boluses to the PTV, apertures to the PTV and boluses to the CTV, and aperture and bolus to the CTV. Results: The planning target volume as specified in ICRU Report 50 can be used only to design the lateral margins of beams, because the distal and proximal margins resulting from CT number uncertainty, beam range uncertainty, tissue motions, and setup uncertainties, are different than the lateral margins resulting from these same factors. The best strategy for target coverage with the planning tools available overirradiated some normal tissues unnecessarily. The available tools also made the planning of lung tumors difficult. Conclusions: This study demonstrated that inclusion of target motion and setup uncertainties into a plan should be performed in the beam design step instead of creating new targets. New computerized treatment planning system tools suggested by this study will ease planning, facilitate abandonment of the PTV concept, improve conformance of the dose distribution to the target, and improve conformal avoidance of critical normal tissues. © 2001 Elsevier Science Inc. Proton, Lung, Target volumes.
INTRODUCTION Standard radiation therapy for patients with inoperable non– small-cell lung carcinoma has been primarily limited by damage to the normal lung tissue that is traversed by the incident X-ray beams. Emami et al. (1) have estimated the TD50/5 for pneumonitis after irradiation of one-third of the lung at 65 Gy for a conventional fractionation rate of 1.8 –2 Gy per day given 5 days per week. Radiation Therapy Oncology Group (RTOG) protocol 93-11 estimates a threshold dose for damage to lung tissue at about 20 Gy, again with standard fractionation (2). Grade 3–5 pneumonitis complications have occurred when this dose was delivered to 50% of the total lung volume or 35 Gy delivered to 60% of the ipsilateral lung. Additional limitations to dose delivery occur when the tumor is in the proximity of the heart, trachea, esophagus, or spinal cord. Proton beams, with their user-selectable range in tissue, can be used to minimize the dose to these normal tissues and possibly increase the dose to the tumor. In May 1995, treatments of inoperable non–small-cell
lung carcinoma patients with proton beams commenced at Loma Linda University Medical Center using a two-arm “in-house” protocol (3,4). Patients were assigned, not randomized, to one of the two arms dependent upon their medical condition. Patients in the first arm received 51 cobalt Gray equivalent (CGE) in 10 fractions, these patients had no evidence of nodal disease and a forced expiratory volume of less than or equal to 1.0 liter. The second arm included patients with nodal disease and a forced expiratory volume greater than 1.0 liter. These patients received 45 CGE in 25 fractions with X-rays to the tumor and mediastinal lymph nodes with a concomitant boost, starting in week two, of 28.8 CGE in 16 fractions with protons. As of July 1998, over 50 patients had received treatment according to this protocol. Beam configurations were devised for these patients, as well as for other patients not on the protocol, with a number of different methods using two different computer systems having different treatment planning tools. These methods can be grouped into three basic planning strategies. This report describes these planning
Reprint requests to: Dr. Michael F. Moyers, Department of Radiation Medicine, 11234 Anderson Street, Loma Linda, California 92354 USA. E-mail: [email protected]
Presented at the AAPM Annual Meeting, San Antonio, Texas, August 9 –13, 1998. Accepted for publication 15 November 2000. 1429
1430
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 1. A typical 3-field beam arrangement for treating lung tumors with proton beams. The red lines are the projections of the aperture edges. The inner red contour is the clinical target volume and the yellow contour is the planning target volume.
strategies and offers a recommendation for a preferred strategy. In addition, suggestions for new computerized treatment-planning system (TPS) tools are presented. METHODS AND MATERIALS In 1993, the International Commission on Radiation Units released a report entitled “Prescribing, Recording, and Reporting Photon Beam Therapy” (5). This report is generally referred to as ICRU 50. The guidelines given in ICRU 50 for defining patient volumes are gradually being accepted throughout the radiotherapy community for photon and neutron beam treatments and are sometimes extended to electron and proton treatments as well. The gross tumor volume (GTV) is defined as the palpable or visible extent of the tumor. The clinical target volume (CTV) is a volume surrounding the GTV that includes that volume with a probable high density of tumor cells due to spread or subclinical disease. The planning target volume (PTV) is a nonanatomical volume that is used to select beam configurations, taking into account geometrical variations such as target motion and setup uncertainty. Neither the width of the lateral penumbra nor the depth dose distribution of the beam are included in the generation of the defined PTV. In 1997, Ekberg et al. (6) published a report that described how ICRU 50 recommended volumes could be applied to treat-
ments of lung tumors with photon beams. That report included the concept of an internal motion margin (IM), which is used to derive an internal target volume (ITV), and a setup margin (SM), used to derive the PTV. The ITV is identical to the mobile target volume (MTV) earlier described by Urie et al. (7) The nomenclature used in this report follows ICRU 50. In this study of treatment planning for proton beam irradiation of lung cancer, one or more of three different strategies were used for each patient. In each case, an attempt was made to follow the ICRU target volume and margin definitions. The first five steps in the planning process were common for each of the strategies. First, the patient was immobilized in a full-body half-cylinder plastic pod with the space between the pod and the patient being filled with a low-density foam. Next, a planning CT consisting of 99 slices was performed and the data transferred to one of the treatment planning workstations. The physician then drew the GTV and devised a surrounding CTV. Finally, the patient was taken to the simulator for a fluoroscopic examination to determine the extent of tumor movement during normal respiration in three perpendicular directions. Both anterior and lateral views were examined and the delineator wires placed at the fullest extent of the tumor motion. Films were then exposed and developed and the motion calculated from the shadows of the delineator wires.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1431
Fig. 2. The dose distribution resulting from the beam arrangement shown in figure 1 illustrating avoidance of the spinal cord, esophagus, and contra-lateral lung. The clinical target volume is shown in white. A rainbow color wash scheme is used to display the dose distribution with the pink/red area representing 90 –100% and the violet/blue area representing 10 –20% of the prescribed dose.
The motion distances were then relayed to the treatment planner.
PTV ⬅ CTV ⫹ IM ⫹ SM
Strategy 1 Strategy 1 is a strategy commonly used for photon planning. In this strategy, a PTV was drawn around the CTV to accommodate both the expected internal target motion within the patient and the setup uncertainty. For these lung tumors, three different motion uncertainties were used for the three transverse directions; rotational motions were not considered. Using a conservative assumption that the set-up uncertainties could be systematic, the motions and setup uncertainties were added linearly with the intent of covering 100% of the CTV by 90% of the prescribed dose under 87% (1.5 ) of all possible conditions. The PTV was created in a single step using an “ellipsoidal expansion” type algorithm where the major radii were the sums of the motion and setup uncertainties in each direction. The PTV can thus be represented by Eq. 1.
The aperture edges for all proton beams were automatically drawn by a TPS tool to project outside of the PTV by a distance input by the planner corresponding to the 90 – 50% penumbra width at the widest extents of the target. Equation 2 describes this lateral aperture margin, AM. Figure 1 is a picture of a typical 3-field arrangement around a PTV. AM around PTV ⬅ (90 –50% penumbra)
(1)
(2)
The bolus thickness distribution was initially designed along diverging ray lines from the source to place the 90% dose on the distal side of the PTV. The distal margin, DM, beyond the target volume was, therefore, zero as given by Eq. 3.
1432
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 3. The dose–volume histogram for the ipsilateral lung of the plan shown in figures 1 and 2. The horizontal axis is the dose in cobalt Gray equivalent (CGE) while the vertical axis is the percentage of the ipsilateral lung volume. The Radiation Therapy Oncology Group (RTOG) 93-11 dose-volume estimate for Grade 3–5 pneumonitis at conventional fractionation schemes is indicated by an asterisk.
DM on PTV ⬅ 0
(3)
The bolus thickness distribution was then expanded laterally (made thinner) to ensure that full lateral proton scatter was achieved (8,9). The lateral distance over which the bolus was expanded, BE, was equal to 3% of the sum of the bolus thickness and the water-equivalent pathlength to the distal side of the PTV. This relationship is given by Eq. 4. BE ⬅ 0.03 䡠 (distal PTV depth ⫹ bolus thickness)
The PTV of Strategy 2 was thus the result of two sequential ellipsoidal expansions. The apertures were designed identically to Strategy 1, as given by Eq. 2. The bolus was designed to the CTV using a custom distal margin that included a 3.5% uncertainty for CT number accuracy and conversion to proton relative linear stopping power and a 3-mm beam range uncertainty for accelerator energy, variable scattering system thickness, bolus density, possibility of bolus voids, and bolus construction. The distal margin on the CTV is given by Eq. 8.
(4) The extent of energy (range) modulation was determined by placing the 90% dose on the proximal side of the PTV using a proximal margin, PM, of zero, as expressed by Eq. 5. PM on PTV ⬅ 0
(5)
Strategy 2 In Strategy 2, an ITV was first designed by adding the internal target motion in each transverse direction to the CTV. ITV ⬅ CTV ⫹ IM
(6)
A PTV was then designed by adding the setup margin to the previously designed ITV. PTV ⬅ ITV ⫹ SM
(7)
DM on CTV ⬅ 0.035 䡠 (distal CTV depth) ⫹ 3 mm (8) The bolus was then expanded laterally by a value given by Eq. 9 to account for internal motion of the CTV, setup uncertainty, and to ensure full lateral proton scatter. The IM used in the bolus expansion calculation was the maximum of the motions perpendicular to the beam direction. BE ⬅ {(IM ⫹ SM)2 ⫹ [0.03 䡠 (distal CTV depth ⫹ bolus thickness)]2}0.5
(9)
The modulation was determined by placing the 90% dose on the proximal side of the CTV using a proximal margin that included a 3.5% CT uncertainty and a 3-mm range uncertainty as expressed by Eq. 10.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1433
Fig. 4. A sagittal reconstruction of a patient scan showing the clinical target volume in red, the derived internal target volume in orange, and the derived planning target volume (PTV) in blue. The yellow contour is the PTV derived using a single derivation that combined the motion and setup uncertainty before expansion. No significant difference between the two methods of derivation is observed.
PM on CTV ⬅ 0.035 䡠 (proximal CTV depth) ⫹ 3 mm (10) Strategy 3 In Strategy 3, no PTV was drawn. The aperture edges for all proton beams were drawn to project outside of the CTV by a distance corresponding to the internal target motion plus the setup uncertainty plus the 90 –50% penumbra width determined at the widest extents of the target. AM around CTV ⬅ IM ⫹ SM ⫹ (90 –50% penumbra) (11) The bolus was designed and expanded laterally identically to Strategy 2 (see Eq. 9). The proximal margin for use in determining the modulation was also identical to Strategy 2 (see Eq. 10). In all three strategies, the dose was prescribed to the isocenter location. Dose estimation calculations were performed with normal range (3 mm), intensity (2%), and CT uncertainties (3.5%).
RESULTS AND DISCUSSION The CTV was identical to the GTV for most patients entered into the protocol study. Most of the tumors in this study were small enough that several CT slices were taken during each breath with little apparent motion of the tumor between adjacent CT scan images. Fluoroscopy yielded typical maximum motions of the tumor of 10 –25 mm in the superior/inferior direction, 5–10 mm in the right/left direction, and 5–10 mm in the anteroposterior direction. The procedure for aligning the patient each day involves comparing a perpendicular set of radiographs to a set of digitally reconstructed radiographs from the treatment planning system. Historically, this procedure has yielded a 5-mm uncertainty of setup. During beam design, a standard setup uncertainty of 5 mm was thus entered for all patients. Typical bolus expansion distances were between 15 and 20 mm for Strategies 2 and 3. The typical beam approach utilized 3 coplanar fields, treating only 2 fields per fraction. Figure 2 is a typical plan illustrating avoidance of the esophagus and spinal cord. Figure 3 is a dose–volume histogram for the same patient’s ipsilateral lung showing
1434
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 5. Coverage of clinical target volume (CTV) using standard bolus design methods is observed to be inadequate when target motion is combined with heterogeneities. The CTV is shown in white and the planning target volume shown in yellow. Nominal doses are shown.
that the dose is well below the level indicated to induce serious pneumonitis using conventional fractionation schemes. Two different computerized TPS were used during the course of this study. Drawing of the PTV for planning using Strategy 1 was difficult using one of the systems because it lacked an automatic region derivation tool to expand the target for motion and setup uncertainty. Drawing of the PTV using the second TPS was rather simple as it allowed derivation of one region from another with expansions in six directions, anterior, posterior, right, left, superior, and inferior. The dose distributions computed with both systems using Strategy 1 showed that the distal 90% dose for individual beams was either too deep or too shallow because the CT and range uncertainties did not match the volumes (PTVs) drawn to indicate motion and setup uncertainty of the target. Strategy 1 was not used for patient treatment. Strategy 2 was confusing for both the planners and physicians because the apertures were designed to one target, the PTV, while the boluses were designed to another, the
CTV. The use of two different targets for a single beam also limited the use of automatic planning tools on both planning systems. This issue was often sidestepped by first designing both the aperture and bolus to the PTV, setting the aperture design on manual, and then redesigning the bolus “automatically” to the CTV. Figure 4 is a sagittal reconstruction showing a CTV drawn in red, the derived ITV in orange, and the derived PTV in blue as given by Eq. 7. The yellow contour is the result of adding the IM and SM together first and then deriving the PTV directly as in Eq. 1. Apparently, the ellipsoidal expansion algorithm used in the region derivation tool does not obey the associative law when used with the voxel sizes encountered with these lung patients. The small differences were not, however, significant to the treatment plans. Another issue with Strategy 2 was the large distance between the CTV used for the bolus design and the aperture edge designed around the PTV. Because the bolus thickness is not defined outside the projection of the target, the bolus design algorithm extrapolates to the aperture edge by using
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1435
Fig. 6. Same plan as in Fig. 5 but with manually designed bolus target shown in blue. Adequate coverage of the clinical target volume by the dose distribution is now provided. Nominal doses are shown.
the thickness of the bolus at the projected edge of the target. This algorithm became problematic, however, in several cases where the integral depth changed rapidly laterally to the CTV. In Fig. 5, an example case is shown where, in the region lateral to the target, the beam travels through less low-density lung but more chest wall and high-density bone. The bolus thickness designed in this region is too thick to allow protons initially travelling lateral to the projection of the target to penetrate and scatter into the target. This results in a decreased number of protons at the target edge and thus, inadequate dose coverage. For this and other cases, the planner manually drew an additional target, to which the bolus was designed. Figure 6 shows how this bolus design target provided adequate coverage of the target without creating overshoot. Strategy 3 had the same issues with bolus design as did Strategy 2. In addition, the planner had to calculate manually the target motions perpendicular to the beam axis and include them into the aperture margins. On the other hand, there was less confusion because there was only one target volume for all beams and beam design steps. Drawing of the
aperture edge when the motions were different in different directions was difficult with one of the planning systems, because it lacked a margining tool that could be varied during the drawing process. A new version of this TPS now has a manual margining tool in which the margin can be changed during drawing. One of the planning systems had the capability of estimating possible doses if the beam was displaced laterally with respect to the patient, the CT numbers or CT number to proton stopping power was incorrect, the beam range was incorrect, or the calibration factor, depth dose, or off-center ratio was incorrect (10,11). Required inputs for this feature (given in the Methods and Materials section) were estimates of the uncertainties. With planning strategies 2 and 3, the “least possible dose” or “minimum dose” to the periphery of the stationary CTV was always 90% or greater except for those cases where an organ at risk was specifically protected. This result was, of course, expected, as the uncertainties were designed into the margins. This uncertainty analysis feature was most useful when optimizing beam arrangements with respect to reducing potential dose errors
1436
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 7. Dose distribution from a single lateral beam showing over-irradiation of normal lung tissue on the opposite side of the patient because of uniform expansion of bolus to account for superior/inferior target motion.
for those cases with nearby heterogeneities or organs at risk. Unfortunately, no planning system currently has the capability of routinely calculating potential doses when one tissue moves with respect to another. Strategies 2 and 3 give identical dose distributions when performed correctly. Strategy 3 was the most desirable method of planning and was used most often for the TPS with a variable distance margining tool. Strategy 2 was used most often for the TPS with the region derivation tool. Approximately one-half of the patients were planned on each TPS. Neither strategy or TPS was ideal so a new TPS tool was developed to design apertures that include directionally dependent target motions and beam set-up uncertainties. For some plans, close visual inspection of the individual beam dose distributions revealed normal tissues that received unnecessary irradiation due to overly expanded boluses. This occurred because the current algorithms force the expansion to be identical in all directions (8,9). Figure 7 shows an example case where dose from a lateral field leaks over to the contralateral lung. In this case, the bolus expansion parameter was set at 16 mm to accommodate the
superior/inferior motion, but only 8 mm was necessary in the anteroposterior direction. When the beam was redesigned with the appropriate value for the anteroposterior direction, the protons stopped short of the lung and the dose to the cord and aorta were reduced (see Fig. 8). A new bolus expansion algorithm was therefore developed to allow directionally dependent values that are automatically keyed to the specified target motions. The problem stated above with having an undefined bolus thickness between the CTV and the aperture edge could be reduced by reducing the motion of the target with respect to the beam during all beam delivery times. This would also dramatically decrease the volume of normal tissues treated. There are several groups studying the feasibility of gated beam delivery with respect to respiratory motion (12). At least one group is studying a method that moves the patient positioner while the beam is on to compensate for intratreatment organ motions (13). Yet another group is taking the approach of controlling the breathing (14). Current research at LLUMC is focussed at combining breathing control with gating.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1437
Fig. 8. Same plan as in Fig. 7 but with the bolus expansion directionally dependent upon the target motion. Irradiation of normal tissues is reduced while maintaining clinical target volume coverage.
Clinical results of the protocol patients treated in this study has been discussed elsewhere (4,15). Briefly, these patients showed a local control rate at 2 years of 87% compared to published data with X-ray beams of 45–58% for comparable types of patients (4). Despite the large margins placed around the CTV to account for motion, high resolution CT scans showed pulmonary injury to be minimal (15). These encouraging results has prompted the dose to be escalated to 6.0 CGE per fraction with a total dose of 60 CGE. CONCLUSION This study of lung tumor treatment planning proved to be an excellent testbed for beam design methodologies and tools. Several new or enhanced tools were developed in response to the study findings. These tools included (1) manual aperture margining with a margin that can be varied during drawing, (2) automatic region derivation with expansion in six directions, (3) automatic aperture margining that
includes directionally dependent target motion and setup uncertainties, and (4) automatic bolus expansion that includes directionally dependent target motion and set-up uncertainties. The first three tools would be of value, not only for charged particle beams but, for X-ray beams as well. With these new TPS tools, the use of the PTV concept can and should be abandoned for charged particle beams. The definition of PTV was deemed inadequate for charged-particle beam planning because the dose with depth changes dramatically just distal to the target. The criteria for expanding the CTV to the PTV for one beam direction may be based on the expected target motion, while for another beam direction will be based on CT number and beam range uncertainties. Incorporating tumor and critical tissue motions and patient setup uncertainties into the beam design step yields better target conformance and critical structure conformal avoidance than designing beams to a motion and setup expanded clinical target volume, i.e., PTV. It is interesting to note
1438
I. J. Radiation Oncology
●
Biology
●
Physics
that several groups have also recommended abandonment of the PTV concept for photon beams (16 –18). The new
Volume 49, Number 5, 2001
TPS tools make the planning of lung tumors with proton beams much simpler and routine.
REFERENCES 1. Emami B, Lyman J, Brown A, et al. Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21:109 –122. 2. Radiation Therapy Oncology Group. A phase I/II dose escalation study using three dimensional conformal radiation therapy post induction chemotherapy in patients with locally advanced inoperable, non-small cell lung cancer. Philadelphia; RTOG 93–11 1994; 4. 3. Slater JM, Slater JD, Bonnet RB. Phase I/II study of combined photon and proton beam therapy for resectable but medically inoperable non-small cell lung cancer. Loma Linda, CA: Loma Linda University Medical Center; 1995. 4. Bush DA, Slater JD, Bonnet R, et al. Proton-beam radiotherapy for early-stage lung cancer. Chest 1999;119:1313–1319. 5. International Commission on Radiation Units and Measurements. Bethesda, MD: ICRU Report 50, Prescribing, Recording, and Reporting Photon Beam Therapy; 1993. 6. Ekberg L, Homberg O, Wittgren L, et al. Application of ICRU Report 50. Bethesda, MD: ICRU News December 1997;4 –9. 7. Urie MM, Goitein M, Doppke K, et al. The role of uncertainty analysis in treatment planning. Int J Radiat Oncol Biol Phys 1991;21(1):91–107. 8. Urie M, Goitein M, Wagner M. Compensating for heterogeneities in proton therapy. Phys Med Biol 1983;29(5):553–566. 9. Goitein M. Compensation for inhomogeneities in charged particle radiotherapy using computed tomography. Int J Radiat Oncol Biol Phys 1978;4:499 –508.
10. Goitein M. Limitations of two-dimensional treatment planning programs. Med Phys 1982;9(4):580 –586. 11. Goitein M. Calculation of the uncertainty in the dose delivered during radiation therapy. Med Phys 1985;12(5):608 – 612. 12. Minohara S, Noda K, Torikoshi M, et al. Irradiation system for heavy ion beams coincident with a patient’s respiratory motion. Particles 1996;17:A19. 13. Morrill SM, Langer M, Lane RG. Real-time couch compensation for intra-treatment organ motion: theoretical advantages. Med Phys 1996;23(6):1083. 14. Wong JW, Sharpe MB, Jaffray DA, et al. The use of active breathing control (ABC) to minimize breathing motion during radiation therapy. Int J Radiat Oncol Biol Phys 1997;39(S2): 164. 15. Bush DA, Dunbar RD, Bonnet R, et al. Pulmonary injury from proton and conventional radiotherapy as revealed by CT. Am J Roentgenol 1999;172(3):735–739. 16. Shalev S. Beam margins and PTV in the presence of organ movement. Med Phys 1995;22(6):938 –939. 17. Fontenla R, Pelizzari CA, Chen GTY. Implications of 3-dimensional target shape and motion in aperture design. Med Phys 1996;23(3);1431–1441. 18. Ten Haken R. Toward dose calculations that include the effects of patient setup uncertainties and organ motion. Med Phys 1998;25(7);A136.
PII S0360-3016(00)01555-8
PHYSICS CONTRIBUTION
METHODOLOGIES AND TOOLS FOR PROTON BEAM DESIGN FOR LUNG TUMORS MICHAEL F. MOYERS, PH.D., DANIEL W. MILLER, PH.D., DAVID A. BUSH, M.D., JERRY D. SLATER, M.D.
AND
Department of Radiation Medicine, Loma Linda University Medical Center, Loma Linda, CA Purpose: Proton beams can potentially increase the dose delivered to lung tumors without increasing the dose to critical normal tissues because protons can be stopped before encountering the normal tissues. This potential can only be realized if tissue motion and planning uncertainties are correctly included during planning. This study evaluated several planning strategies to determine which method best provides adequate tumor coverage, minimal normal tissue irradiation, and simplicity of use. Methods and Materials: Proton beam treatment plans were generated using one or more of three different planning strategies. These strategies included designing apertures and boluses to the PTV, apertures to the PTV and boluses to the CTV, and aperture and bolus to the CTV. Results: The planning target volume as specified in ICRU Report 50 can be used only to design the lateral margins of beams, because the distal and proximal margins resulting from CT number uncertainty, beam range uncertainty, tissue motions, and setup uncertainties, are different than the lateral margins resulting from these same factors. The best strategy for target coverage with the planning tools available overirradiated some normal tissues unnecessarily. The available tools also made the planning of lung tumors difficult. Conclusions: This study demonstrated that inclusion of target motion and setup uncertainties into a plan should be performed in the beam design step instead of creating new targets. New computerized treatment planning system tools suggested by this study will ease planning, facilitate abandonment of the PTV concept, improve conformance of the dose distribution to the target, and improve conformal avoidance of critical normal tissues. © 2001 Elsevier Science Inc. Proton, Lung, Target volumes.
INTRODUCTION Standard radiation therapy for patients with inoperable non– small-cell lung carcinoma has been primarily limited by damage to the normal lung tissue that is traversed by the incident X-ray beams. Emami et al. (1) have estimated the TD50/5 for pneumonitis after irradiation of one-third of the lung at 65 Gy for a conventional fractionation rate of 1.8 –2 Gy per day given 5 days per week. Radiation Therapy Oncology Group (RTOG) protocol 93-11 estimates a threshold dose for damage to lung tissue at about 20 Gy, again with standard fractionation (2). Grade 3–5 pneumonitis complications have occurred when this dose was delivered to 50% of the total lung volume or 35 Gy delivered to 60% of the ipsilateral lung. Additional limitations to dose delivery occur when the tumor is in the proximity of the heart, trachea, esophagus, or spinal cord. Proton beams, with their user-selectable range in tissue, can be used to minimize the dose to these normal tissues and possibly increase the dose to the tumor. In May 1995, treatments of inoperable non–small-cell
lung carcinoma patients with proton beams commenced at Loma Linda University Medical Center using a two-arm “in-house” protocol (3,4). Patients were assigned, not randomized, to one of the two arms dependent upon their medical condition. Patients in the first arm received 51 cobalt Gray equivalent (CGE) in 10 fractions, these patients had no evidence of nodal disease and a forced expiratory volume of less than or equal to 1.0 liter. The second arm included patients with nodal disease and a forced expiratory volume greater than 1.0 liter. These patients received 45 CGE in 25 fractions with X-rays to the tumor and mediastinal lymph nodes with a concomitant boost, starting in week two, of 28.8 CGE in 16 fractions with protons. As of July 1998, over 50 patients had received treatment according to this protocol. Beam configurations were devised for these patients, as well as for other patients not on the protocol, with a number of different methods using two different computer systems having different treatment planning tools. These methods can be grouped into three basic planning strategies. This report describes these planning
Reprint requests to: Dr. Michael F. Moyers, Department of Radiation Medicine, 11234 Anderson Street, Loma Linda, California 92354 USA. E-mail: [email protected]
Presented at the AAPM Annual Meeting, San Antonio, Texas, August 9 –13, 1998. Accepted for publication 15 November 2000. 1429
1430
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 1. A typical 3-field beam arrangement for treating lung tumors with proton beams. The red lines are the projections of the aperture edges. The inner red contour is the clinical target volume and the yellow contour is the planning target volume.
strategies and offers a recommendation for a preferred strategy. In addition, suggestions for new computerized treatment-planning system (TPS) tools are presented. METHODS AND MATERIALS In 1993, the International Commission on Radiation Units released a report entitled “Prescribing, Recording, and Reporting Photon Beam Therapy” (5). This report is generally referred to as ICRU 50. The guidelines given in ICRU 50 for defining patient volumes are gradually being accepted throughout the radiotherapy community for photon and neutron beam treatments and are sometimes extended to electron and proton treatments as well. The gross tumor volume (GTV) is defined as the palpable or visible extent of the tumor. The clinical target volume (CTV) is a volume surrounding the GTV that includes that volume with a probable high density of tumor cells due to spread or subclinical disease. The planning target volume (PTV) is a nonanatomical volume that is used to select beam configurations, taking into account geometrical variations such as target motion and setup uncertainty. Neither the width of the lateral penumbra nor the depth dose distribution of the beam are included in the generation of the defined PTV. In 1997, Ekberg et al. (6) published a report that described how ICRU 50 recommended volumes could be applied to treat-
ments of lung tumors with photon beams. That report included the concept of an internal motion margin (IM), which is used to derive an internal target volume (ITV), and a setup margin (SM), used to derive the PTV. The ITV is identical to the mobile target volume (MTV) earlier described by Urie et al. (7) The nomenclature used in this report follows ICRU 50. In this study of treatment planning for proton beam irradiation of lung cancer, one or more of three different strategies were used for each patient. In each case, an attempt was made to follow the ICRU target volume and margin definitions. The first five steps in the planning process were common for each of the strategies. First, the patient was immobilized in a full-body half-cylinder plastic pod with the space between the pod and the patient being filled with a low-density foam. Next, a planning CT consisting of 99 slices was performed and the data transferred to one of the treatment planning workstations. The physician then drew the GTV and devised a surrounding CTV. Finally, the patient was taken to the simulator for a fluoroscopic examination to determine the extent of tumor movement during normal respiration in three perpendicular directions. Both anterior and lateral views were examined and the delineator wires placed at the fullest extent of the tumor motion. Films were then exposed and developed and the motion calculated from the shadows of the delineator wires.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1431
Fig. 2. The dose distribution resulting from the beam arrangement shown in figure 1 illustrating avoidance of the spinal cord, esophagus, and contra-lateral lung. The clinical target volume is shown in white. A rainbow color wash scheme is used to display the dose distribution with the pink/red area representing 90 –100% and the violet/blue area representing 10 –20% of the prescribed dose.
The motion distances were then relayed to the treatment planner.
PTV ⬅ CTV ⫹ IM ⫹ SM
Strategy 1 Strategy 1 is a strategy commonly used for photon planning. In this strategy, a PTV was drawn around the CTV to accommodate both the expected internal target motion within the patient and the setup uncertainty. For these lung tumors, three different motion uncertainties were used for the three transverse directions; rotational motions were not considered. Using a conservative assumption that the set-up uncertainties could be systematic, the motions and setup uncertainties were added linearly with the intent of covering 100% of the CTV by 90% of the prescribed dose under 87% (1.5 ) of all possible conditions. The PTV was created in a single step using an “ellipsoidal expansion” type algorithm where the major radii were the sums of the motion and setup uncertainties in each direction. The PTV can thus be represented by Eq. 1.
The aperture edges for all proton beams were automatically drawn by a TPS tool to project outside of the PTV by a distance input by the planner corresponding to the 90 – 50% penumbra width at the widest extents of the target. Equation 2 describes this lateral aperture margin, AM. Figure 1 is a picture of a typical 3-field arrangement around a PTV. AM around PTV ⬅ (90 –50% penumbra)
(1)
(2)
The bolus thickness distribution was initially designed along diverging ray lines from the source to place the 90% dose on the distal side of the PTV. The distal margin, DM, beyond the target volume was, therefore, zero as given by Eq. 3.
1432
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 3. The dose–volume histogram for the ipsilateral lung of the plan shown in figures 1 and 2. The horizontal axis is the dose in cobalt Gray equivalent (CGE) while the vertical axis is the percentage of the ipsilateral lung volume. The Radiation Therapy Oncology Group (RTOG) 93-11 dose-volume estimate for Grade 3–5 pneumonitis at conventional fractionation schemes is indicated by an asterisk.
DM on PTV ⬅ 0
(3)
The bolus thickness distribution was then expanded laterally (made thinner) to ensure that full lateral proton scatter was achieved (8,9). The lateral distance over which the bolus was expanded, BE, was equal to 3% of the sum of the bolus thickness and the water-equivalent pathlength to the distal side of the PTV. This relationship is given by Eq. 4. BE ⬅ 0.03 䡠 (distal PTV depth ⫹ bolus thickness)
The PTV of Strategy 2 was thus the result of two sequential ellipsoidal expansions. The apertures were designed identically to Strategy 1, as given by Eq. 2. The bolus was designed to the CTV using a custom distal margin that included a 3.5% uncertainty for CT number accuracy and conversion to proton relative linear stopping power and a 3-mm beam range uncertainty for accelerator energy, variable scattering system thickness, bolus density, possibility of bolus voids, and bolus construction. The distal margin on the CTV is given by Eq. 8.
(4) The extent of energy (range) modulation was determined by placing the 90% dose on the proximal side of the PTV using a proximal margin, PM, of zero, as expressed by Eq. 5. PM on PTV ⬅ 0
(5)
Strategy 2 In Strategy 2, an ITV was first designed by adding the internal target motion in each transverse direction to the CTV. ITV ⬅ CTV ⫹ IM
(6)
A PTV was then designed by adding the setup margin to the previously designed ITV. PTV ⬅ ITV ⫹ SM
(7)
DM on CTV ⬅ 0.035 䡠 (distal CTV depth) ⫹ 3 mm (8) The bolus was then expanded laterally by a value given by Eq. 9 to account for internal motion of the CTV, setup uncertainty, and to ensure full lateral proton scatter. The IM used in the bolus expansion calculation was the maximum of the motions perpendicular to the beam direction. BE ⬅ {(IM ⫹ SM)2 ⫹ [0.03 䡠 (distal CTV depth ⫹ bolus thickness)]2}0.5
(9)
The modulation was determined by placing the 90% dose on the proximal side of the CTV using a proximal margin that included a 3.5% CT uncertainty and a 3-mm range uncertainty as expressed by Eq. 10.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1433
Fig. 4. A sagittal reconstruction of a patient scan showing the clinical target volume in red, the derived internal target volume in orange, and the derived planning target volume (PTV) in blue. The yellow contour is the PTV derived using a single derivation that combined the motion and setup uncertainty before expansion. No significant difference between the two methods of derivation is observed.
PM on CTV ⬅ 0.035 䡠 (proximal CTV depth) ⫹ 3 mm (10) Strategy 3 In Strategy 3, no PTV was drawn. The aperture edges for all proton beams were drawn to project outside of the CTV by a distance corresponding to the internal target motion plus the setup uncertainty plus the 90 –50% penumbra width determined at the widest extents of the target. AM around CTV ⬅ IM ⫹ SM ⫹ (90 –50% penumbra) (11) The bolus was designed and expanded laterally identically to Strategy 2 (see Eq. 9). The proximal margin for use in determining the modulation was also identical to Strategy 2 (see Eq. 10). In all three strategies, the dose was prescribed to the isocenter location. Dose estimation calculations were performed with normal range (3 mm), intensity (2%), and CT uncertainties (3.5%).
RESULTS AND DISCUSSION The CTV was identical to the GTV for most patients entered into the protocol study. Most of the tumors in this study were small enough that several CT slices were taken during each breath with little apparent motion of the tumor between adjacent CT scan images. Fluoroscopy yielded typical maximum motions of the tumor of 10 –25 mm in the superior/inferior direction, 5–10 mm in the right/left direction, and 5–10 mm in the anteroposterior direction. The procedure for aligning the patient each day involves comparing a perpendicular set of radiographs to a set of digitally reconstructed radiographs from the treatment planning system. Historically, this procedure has yielded a 5-mm uncertainty of setup. During beam design, a standard setup uncertainty of 5 mm was thus entered for all patients. Typical bolus expansion distances were between 15 and 20 mm for Strategies 2 and 3. The typical beam approach utilized 3 coplanar fields, treating only 2 fields per fraction. Figure 2 is a typical plan illustrating avoidance of the esophagus and spinal cord. Figure 3 is a dose–volume histogram for the same patient’s ipsilateral lung showing
1434
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 5. Coverage of clinical target volume (CTV) using standard bolus design methods is observed to be inadequate when target motion is combined with heterogeneities. The CTV is shown in white and the planning target volume shown in yellow. Nominal doses are shown.
that the dose is well below the level indicated to induce serious pneumonitis using conventional fractionation schemes. Two different computerized TPS were used during the course of this study. Drawing of the PTV for planning using Strategy 1 was difficult using one of the systems because it lacked an automatic region derivation tool to expand the target for motion and setup uncertainty. Drawing of the PTV using the second TPS was rather simple as it allowed derivation of one region from another with expansions in six directions, anterior, posterior, right, left, superior, and inferior. The dose distributions computed with both systems using Strategy 1 showed that the distal 90% dose for individual beams was either too deep or too shallow because the CT and range uncertainties did not match the volumes (PTVs) drawn to indicate motion and setup uncertainty of the target. Strategy 1 was not used for patient treatment. Strategy 2 was confusing for both the planners and physicians because the apertures were designed to one target, the PTV, while the boluses were designed to another, the
CTV. The use of two different targets for a single beam also limited the use of automatic planning tools on both planning systems. This issue was often sidestepped by first designing both the aperture and bolus to the PTV, setting the aperture design on manual, and then redesigning the bolus “automatically” to the CTV. Figure 4 is a sagittal reconstruction showing a CTV drawn in red, the derived ITV in orange, and the derived PTV in blue as given by Eq. 7. The yellow contour is the result of adding the IM and SM together first and then deriving the PTV directly as in Eq. 1. Apparently, the ellipsoidal expansion algorithm used in the region derivation tool does not obey the associative law when used with the voxel sizes encountered with these lung patients. The small differences were not, however, significant to the treatment plans. Another issue with Strategy 2 was the large distance between the CTV used for the bolus design and the aperture edge designed around the PTV. Because the bolus thickness is not defined outside the projection of the target, the bolus design algorithm extrapolates to the aperture edge by using
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1435
Fig. 6. Same plan as in Fig. 5 but with manually designed bolus target shown in blue. Adequate coverage of the clinical target volume by the dose distribution is now provided. Nominal doses are shown.
the thickness of the bolus at the projected edge of the target. This algorithm became problematic, however, in several cases where the integral depth changed rapidly laterally to the CTV. In Fig. 5, an example case is shown where, in the region lateral to the target, the beam travels through less low-density lung but more chest wall and high-density bone. The bolus thickness designed in this region is too thick to allow protons initially travelling lateral to the projection of the target to penetrate and scatter into the target. This results in a decreased number of protons at the target edge and thus, inadequate dose coverage. For this and other cases, the planner manually drew an additional target, to which the bolus was designed. Figure 6 shows how this bolus design target provided adequate coverage of the target without creating overshoot. Strategy 3 had the same issues with bolus design as did Strategy 2. In addition, the planner had to calculate manually the target motions perpendicular to the beam axis and include them into the aperture margins. On the other hand, there was less confusion because there was only one target volume for all beams and beam design steps. Drawing of the
aperture edge when the motions were different in different directions was difficult with one of the planning systems, because it lacked a margining tool that could be varied during the drawing process. A new version of this TPS now has a manual margining tool in which the margin can be changed during drawing. One of the planning systems had the capability of estimating possible doses if the beam was displaced laterally with respect to the patient, the CT numbers or CT number to proton stopping power was incorrect, the beam range was incorrect, or the calibration factor, depth dose, or off-center ratio was incorrect (10,11). Required inputs for this feature (given in the Methods and Materials section) were estimates of the uncertainties. With planning strategies 2 and 3, the “least possible dose” or “minimum dose” to the periphery of the stationary CTV was always 90% or greater except for those cases where an organ at risk was specifically protected. This result was, of course, expected, as the uncertainties were designed into the margins. This uncertainty analysis feature was most useful when optimizing beam arrangements with respect to reducing potential dose errors
1436
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 49, Number 5, 2001
Fig. 7. Dose distribution from a single lateral beam showing over-irradiation of normal lung tissue on the opposite side of the patient because of uniform expansion of bolus to account for superior/inferior target motion.
for those cases with nearby heterogeneities or organs at risk. Unfortunately, no planning system currently has the capability of routinely calculating potential doses when one tissue moves with respect to another. Strategies 2 and 3 give identical dose distributions when performed correctly. Strategy 3 was the most desirable method of planning and was used most often for the TPS with a variable distance margining tool. Strategy 2 was used most often for the TPS with the region derivation tool. Approximately one-half of the patients were planned on each TPS. Neither strategy or TPS was ideal so a new TPS tool was developed to design apertures that include directionally dependent target motions and beam set-up uncertainties. For some plans, close visual inspection of the individual beam dose distributions revealed normal tissues that received unnecessary irradiation due to overly expanded boluses. This occurred because the current algorithms force the expansion to be identical in all directions (8,9). Figure 7 shows an example case where dose from a lateral field leaks over to the contralateral lung. In this case, the bolus expansion parameter was set at 16 mm to accommodate the
superior/inferior motion, but only 8 mm was necessary in the anteroposterior direction. When the beam was redesigned with the appropriate value for the anteroposterior direction, the protons stopped short of the lung and the dose to the cord and aorta were reduced (see Fig. 8). A new bolus expansion algorithm was therefore developed to allow directionally dependent values that are automatically keyed to the specified target motions. The problem stated above with having an undefined bolus thickness between the CTV and the aperture edge could be reduced by reducing the motion of the target with respect to the beam during all beam delivery times. This would also dramatically decrease the volume of normal tissues treated. There are several groups studying the feasibility of gated beam delivery with respect to respiratory motion (12). At least one group is studying a method that moves the patient positioner while the beam is on to compensate for intratreatment organ motions (13). Yet another group is taking the approach of controlling the breathing (14). Current research at LLUMC is focussed at combining breathing control with gating.
Proton beam design for lung tumors
●
M. F. MOYERS et al.
1437
Fig. 8. Same plan as in Fig. 7 but with the bolus expansion directionally dependent upon the target motion. Irradiation of normal tissues is reduced while maintaining clinical target volume coverage.
Clinical results of the protocol patients treated in this study has been discussed elsewhere (4,15). Briefly, these patients showed a local control rate at 2 years of 87% compared to published data with X-ray beams of 45–58% for comparable types of patients (4). Despite the large margins placed around the CTV to account for motion, high resolution CT scans showed pulmonary injury to be minimal (15). These encouraging results has prompted the dose to be escalated to 6.0 CGE per fraction with a total dose of 60 CGE. CONCLUSION This study of lung tumor treatment planning proved to be an excellent testbed for beam design methodologies and tools. Several new or enhanced tools were developed in response to the study findings. These tools included (1) manual aperture margining with a margin that can be varied during drawing, (2) automatic region derivation with expansion in six directions, (3) automatic aperture margining that
includes directionally dependent target motion and setup uncertainties, and (4) automatic bolus expansion that includes directionally dependent target motion and set-up uncertainties. The first three tools would be of value, not only for charged particle beams but, for X-ray beams as well. With these new TPS tools, the use of the PTV concept can and should be abandoned for charged particle beams. The definition of PTV was deemed inadequate for charged-particle beam planning because the dose with depth changes dramatically just distal to the target. The criteria for expanding the CTV to the PTV for one beam direction may be based on the expected target motion, while for another beam direction will be based on CT number and beam range uncertainties. Incorporating tumor and critical tissue motions and patient setup uncertainties into the beam design step yields better target conformance and critical structure conformal avoidance than designing beams to a motion and setup expanded clinical target volume, i.e., PTV. It is interesting to note
1438
I. J. Radiation Oncology
●
Biology
●
Physics
that several groups have also recommended abandonment of the PTV concept for photon beams (16 –18). The new
Volume 49, Number 5, 2001
TPS tools make the planning of lung tumors with proton beams much simpler and routine.
REFERENCES 1. Emami B, Lyman J, Brown A, et al. Tolerance of normal tissue to therapeutic irradiation. Int J Radiat Oncol Biol Phys 1991;21:109 –122. 2. Radiation Therapy Oncology Group. A phase I/II dose escalation study using three dimensional conformal radiation therapy post induction chemotherapy in patients with locally advanced inoperable, non-small cell lung cancer. Philadelphia; RTOG 93–11 1994; 4. 3. Slater JM, Slater JD, Bonnet RB. Phase I/II study of combined photon and proton beam therapy for resectable but medically inoperable non-small cell lung cancer. Loma Linda, CA: Loma Linda University Medical Center; 1995. 4. Bush DA, Slater JD, Bonnet R, et al. Proton-beam radiotherapy for early-stage lung cancer. Chest 1999;119:1313–1319. 5. International Commission on Radiation Units and Measurements. Bethesda, MD: ICRU Report 50, Prescribing, Recording, and Reporting Photon Beam Therapy; 1993. 6. Ekberg L, Homberg O, Wittgren L, et al. Application of ICRU Report 50. Bethesda, MD: ICRU News December 1997;4 –9. 7. Urie MM, Goitein M, Doppke K, et al. The role of uncertainty analysis in treatment planning. Int J Radiat Oncol Biol Phys 1991;21(1):91–107. 8. Urie M, Goitein M, Wagner M. Compensating for heterogeneities in proton therapy. Phys Med Biol 1983;29(5):553–566. 9. Goitein M. Compensation for inhomogeneities in charged particle radiotherapy using computed tomography. Int J Radiat Oncol Biol Phys 1978;4:499 –508.
10. Goitein M. Limitations of two-dimensional treatment planning programs. Med Phys 1982;9(4):580 –586. 11. Goitein M. Calculation of the uncertainty in the dose delivered during radiation therapy. Med Phys 1985;12(5):608 – 612. 12. Minohara S, Noda K, Torikoshi M, et al. Irradiation system for heavy ion beams coincident with a patient’s respiratory motion. Particles 1996;17:A19. 13. Morrill SM, Langer M, Lane RG. Real-time couch compensation for intra-treatment organ motion: theoretical advantages. Med Phys 1996;23(6):1083. 14. Wong JW, Sharpe MB, Jaffray DA, et al. The use of active breathing control (ABC) to minimize breathing motion during radiation therapy. Int J Radiat Oncol Biol Phys 1997;39(S2): 164. 15. Bush DA, Dunbar RD, Bonnet R, et al. Pulmonary injury from proton and conventional radiotherapy as revealed by CT. Am J Roentgenol 1999;172(3):735–739. 16. Shalev S. Beam margins and PTV in the presence of organ movement. Med Phys 1995;22(6):938 –939. 17. Fontenla R, Pelizzari CA, Chen GTY. Implications of 3-dimensional target shape and motion in aperture design. Med Phys 1996;23(3);1431–1441. 18. Ten Haken R. Toward dose calculations that include the effects of patient setup uncertainties and organ motion. Med Phys 1998;25(7);A136.