Patient positioning for protontherapy using a proton ... - Science Direct

2@3

Nuclear Instruments

and Methods in Physics Research A 356 (1995) 558-565

. -9

f!!ll

ELSEVIER

Patient positioning for protontherapy

NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH SectjonA

using a proton range telescope

*,

J.L. Romero a, J.H. Osborne a, F.P. Brady a, W. Caskey a, D.A. Cebra a, M.D. Partlan a, B.H. Kusko b, R.S. King b, I. Mirshad b, H. Kubo ‘ld, I. Daftari cTd,W. Chu e a Department of Physics, University of California, Davis, CA 95616, USA b Cracker Nuclear Laboratory, University of California, Dauis, CA, USA ’ Department of Radiation Oncology, University of California Dacis Medical Center, Daois, CA, LISA d University of California, San Francisco, CA, USA e Lawrence Berkeley Laboratory, Uniuersify of California, Berkeley, CA, USA Received 18 May 1994

Abstract A method is described for imaging integrated density along the beam path in phantoms that makes use of high energy proton beams. An application of the technique is in the positioning of patients in a proton therapy radiation facility. It makes use of a proton range telescope for density variation and X-Y detectors for planar positioning. The principle of the method was tested at a proton energy of 66 MeV. Good visual quality is seen in the tests. The measurements are compared with detailed Monte Carlo simulations, and good agreement is found. We apply the simulations to high energy proton beams (245 MeV) and show that the method should provide good visual quality and sensitivity for positioning at the higher energies necessary for whole body therapy.

1. Introduction Proton radiotherapy (PRT) is a technologically advanced means of treating cancer and other diseases. A high energy proton beam can accurately deliver a precise radiation dose to the diseased volume while minimizing the dose to the surrounding tissue. At present, proton and heavy ion radiotherapy is being used at various centers around the world. The first hospital based proton machine is operating at Loma Linda University medical center in Southern California, USA [l]. The first hospital-based heavy ion machine has been in operation in Chiba, Japan since October, 1993 [2]. In the USA, general clinical specifications for a proton facility have been recommended [3]. Following these recommendations, the construction of a cyclotron capable of accelerating protons up to = 250 MeV has been started at Massachussets General Hospital in Boston, Massachussets. The success of proton radiotherapy is based on the precision with which the dose is deposited in the tumor volume [3]. To take the full advantage of charged particle beams, the accuracy in positioning the patient within the radiation field should be comparable to the precision in obtaining the dose localization. At present 2 to 3 mm

* Corresponding

author.

016%9002/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDI 0168-9002(94)01353-5

accuracy is achieved [4-51 by using rigid immobilization devices and monitoring the body surface with closeup TV cameras. This way of monitoring is often difficult because the immobilization devices attached to the patient do not necessarily accurately reflect the location of internal structures. During the last few years many institutions using conventional radiation therapy have adapted commercial based on-line portal imaging devices to detect errors that occur during the patient treatment [6]. In this paper, we discuss a new method of locating, positioning and monitoring the patient and tumor volume that uses the therapy proton beam per se and techniques of high energy nuclear physics. In Section 2 we review previous work and describe the principle of the method. Section 3 discusses the experimental setup used in these tests at 66 MeV. Results are presented in Section 4 corresponding to some of the phantoms used in these tests and comparisons are made with Monte Carlo simulations. Section 5 presents results from simulations with 245 MeV protons.

2. Principle of the method The value of using beams of high-energy charged particles, in particular proton beams, as sensitive probes to detect small density variations has been recognized for

J.L. Romero et al. /Nucl.

559

Instr. and Meth. in Phys. Res. A 356 (1995) 558-565

many years. In 1968, Koehler used 160 MeV protons from the Harvard University cyclotron to obtain radiography of fabricated phantoms [7]. He showed that density differences as small as 0.05% could be detected. Working with Steward in the early 1970s he obtained proton radiography of real tumors of the breast [8] and brain [9]. In the late 1970s Hanson et al. used a 240 MeV proton beam from the Los Alamos Meson Production Factory (LAMPF) to obtain computed tomography of the highest quality and sensitivity to density variation of the order of 0.013 g/cm2 [lO,ll]. They also compared proton radiography with conventional X-ray radiography and showed that for similar quality images, protons resulted in a much lower dose to the patient than did X-rays, and that treatment doses could be located more accurately and away from sensitive areas. Similar results were obtained by the Berkeley group using 900 MeV alpha particles from the 184 in. synchrocyclotron [12]. A general discussion on the physics of imaging with ionizing radiations is found in Ref. [13]. A good qualitative description of the principle of the method has been given by Tobias et al. [14]: “The dominant factor determining the range of a proton beam of a given energy in a material is the stopping power of the material (stopping power is roughly proportional to mass density). Thus a monoenergetic beam of particles will penetrate an absorber to a well-defined depth. If the initial energy (range) is adjusted to slightly exceed the range in a homogeneous absorber, all particles will completely traverse the absorber and will register in a suitable detector placed immediately downstream of the absorber. When a feature of higher stopping power than that of the absorber in introduced into the material, the range of particles traversing the feature will be less than the range of particles through the surrounding material. If the feature is of lower stopping power relative to the rest of the target, the range will be larger. The precise amount of relative displacement of the nominal stopping power and thickness of the feature”. The quantitative extension of this statement will be addressed shortly, with the important qualification that the precision in the measurement of the integrated density along the beam path is limited by the range straggling, while the lateral precision is limited by multiple scattering. Fig. 1 presents the well-known relationship between the proton energy and the range in several materials including water and also the corresponding range straggling [15]. It is seen that for water, straggling is typically 1% of the range. For a specimen of 40 cm of water, straggling is about 4 mm. Multiple (Coulomb) scattering of the proton beam with the atoms of the material affects the lateral precision. However, a determination of the trajectory of the proton, measured before and after traversing the phantom, allows us to correct for this effect. Typically, one expects to correct the lateral component of the proton trajectory position to within 1 mm [lo].

-Pb - 350..,,,.

-cu ~l,l~~,l,.,I,,,,.,

+ Be .. -water 1.4

4upper CUMS cmespand

I .I. 20

40

11,,

1.2

-

0.2

I 100

60 80 Range (g/cm’)

-

120

0.0

Fig. 1. Proton kinetic energy (left scale) and range straggling (right scale) for Pb, Cu, Be, and water as a function of the range. After Ref. [15].

Considering the above, a conceptual design of a detector array to obtain a radiography of a patient tumor or phantom structure, by measuring the range of a high energy proton beam is sketched in Fig. 2. A (very) low intensity proton beam passes an X-Y position-determining device such as a multiwire proportional chamber, impinges on a patient, traverses a density anomaly such as a tumor or bone structure, emerges through a second X-Y device and finally is stopped in a total energy detector, such as the

16

AE detectors

each 1 mm thick

7

il

-

Wire Chamber 2

Patient

Tumor

lv

Wire Chamber 1

Proton beam (200-300 Mew)

Fig. 2. Conceptual design of a detecting system to obtain a proton “radiography” of a patient tumor or phantom structure. In essence, WC1 and WC2 measure the planar (X-Y) position of the tumor, while the integrated density along the beam path of the tumor (or density variation) is obtained from the range detector. For 245 MeV incident protons, the latter could consists of up to 18 thin AE detectors (1 mm in thickness, for instance), in coincidence with each other and with the wire chambers.

560

J.L. Romero et al. /f&cl. Trigger detector

AEZ

instr. and Metk. in Phys. Res. A 356 (199.5) 558-565

AE4

Proton Beam

I

AEdetectors are plastic scmtillators l-2 mm thick

Degrader Phantom AEl

BE3

AE5

Fig. 3. Experimental setup using protons with au energy of 66.0 MeV at the entrance of the phantom, The phantom shown corresponds to that of Fig. 5.

range telescope. In this case, if one uses 1 mm scintillating detectors, the typical average error in determining the range (or depth sensitivity) is 0.05 g/cm2.

0.8 to 2 mm. An adjustable energy degrader and a trigger AE detector are located before the phantom. Wire chambers WC1 and WC2, located before and after the phantom, respectively, are used to determine more precisely the individual proton trajectory. The multiwire chambers give position (one standard deviation) close to u= f 1 mm [17], and also allow corrections for the effects of multiple Coulomb scattering. The anode signal from each AE detector is fed to a fast discriminator. The threshold for the latter is set in a separate calibration run using full energy protons (67.5 MeV), guaranteeing in this way the detection of low energy protons when the phantom is used. The output from the fast discriminator is fed into a CAMAC coincidence register, which establishes how many of the five AE detectors are in coincidence with the trigger detector. The latter initiates the sequence of electronic coincidences for each event. The coincidence register and the X and Y signals from the wire chambers are stored event-by-event onto magnetic tape for off-line analysis. In this manner, each individual proton is registered as an electronic coincidence between the trigger detector and at least one of the five AE detectors. For example, an event that has a coincidence register set to detectors 1, 2 and 3 is accepted in the analysis. A coincidence register set to detectors 1, 2, 3 and 5 is rejected. In the next section, we present results from events that have consecutive hits starting from detector 1.

3. Experimental setup In these first tests we tried to locate simple structures to demonstrate the feasibility of the method. An important aim of these tests was to compare the experimental results with detailed Monte Carlo simulations. This was done so we may confidently use the simulations at proton energies of around 245 MeV, the proposed energy of the new proton therapy accelerator. In order to investigate the feasibility of such a device, we have performed tests at the Cracker Nuclear Laboratory (CNL) Cyclotron at UC Davis. This cyclotron provides a maximum proton beam energy of 67.5 MeV, with the important characteristic that this direct beam can have an extremely low intensity (100 to 1000 protons/s), obtained with no collimation. Details concerning the production and operating features of this low intensity proton beam are given in Ref. [16]. Fig. 3 shows the experimental setup we used at CNL. Use was made of existing AE detectors and available wire chambers [17] to assemble a fast range telescope. This system differs from the pioneering method used by Hansen et al. [ll], in that this telescope can count much faster (at = lob/s), is inexpensive, and fairly simple to setup, calibrate, and use. The low intensity proton beam impinges on a phantom and onto the detecting telescope, where it stops. The fast range telescope consists of five thin AE detectors (AEl, AE2, . . , AE5), with thicknesses ranging from

4. Results Fig. 4 (right) presents the results corresponding to an acrylic phantom shaped as in Fig. 4 (left). The phantom has, as shown, a raised pattern 1.6 mm in thickness on top of an acrylic cube 30 mm in thickness. The very low intensity proton beam (about 160 protons/s with an energy at the phantom of 66.0 MeV) impinged normal to the phantom along the Z direction. Each panel in Fig. 4 shows the distribution of events stopped in the corresponding AE detector. The X, Y values correspond to the hits in WC1 placed upstream of the phantom, as in the setup in Fig. 3. The sequence of scatter plots of positions at the entrance wire chamber for the cases where the proton stopped in the first, second, . . , and fifth detector is an excellent illustration of both the principle and sensitivity of the method. of the phantom was obtained with This “radiography” about 3 x lo4 protons and it took 3 min. It is seen that most of the protons incident on the thin raised pattern are stopped in AE2, while most of the protons incident outside are stopped in AE4. Straggling, beam energy width ((T = 0.3 MeV), wire chambers resolution (a = 1.3 mm) and multiple scattering produce diffuseness in depth (as measured by the detector where the proton stopped) and in the X, Y position (as measured by the hit in WCl). The rather good depth sensitivity allows us, in the present configura-



56

Data

6

0.16 an slab Acrylic

0 lY

R,

I

I I I

Proton

6

x Beam ‘, Direction

k_--__.x \ \

\

\

\

‘\ 2

AE4

6

AE5

~.1.1....1....11...I....,~

0

2 4 Horizontal Position (cm)

6

Fig. 4. Phantom used (left) for the results presented at the right. The energy of the incident proton beam was 66.0 MeV. The figure on the right shows the pattern observed in WC1 for protons stopped in each AE detector.

562

J.L. Romero et al. / Nucl. Instr. and Meth. in Phys. Res. A 356 (1995) 558-565

tion, to distinguish thickness differences of at least 0.05 g/cm2 (see below). Another set of measurements were made using a variety of simple phantoms. The purpose here was to explore the capability of our simulation code to reproduce the experimentally observed results. Here we discuss the results using the phantom and geometry shown in Fig. 5. Briefly, the Monte Carlo simulations are done as follows: Individual protons, each with an energy of 66.0 MeV,are sent along the beam direction, the Z-axis, at random X, Y points, so they span the phantom. The incident energy is smeared according to a Gaussian distribution corresponding to a beam energy resolution with TV= 0.3 MeV. Each proton is then propagated through the phantom and, incrementally (in the beam direction) the new energy (decreased due to energy loss) and new position due to multiple scattering are calculated. The latter is done in three dimensions, so at each increment in the Z-axis a new position (X, Y, Z) and a new energy is obtained. This process is continued until the proton energy is zero (normally it stops in one of the AE detectors). At the point of stopping, the final position is determined by smearing the stopping position according to the rangestraggling curve, using a Gaussian distribution with a standard deviation obtained from Fig. 1. Finally, the X, Y hits at the location of the wire chambers are also smeared according to the expected resolution of the wire chambers (a = 1.3 mm). The process is repeated for the desired number of protons. See Ref. 1181 for details. Fig. 6 shows X-Y scatter plots as given by WC1 and corresponding to protons stopping in AEl through AE.5. The total number of protons was 6723. On the right in Fig. 6 is shown the corresponding Monte Carlo simulation. Good overall visual agreement is observed between the

0.32 +,‘,‘e

0.16 +,:c

h’

0.95 cm :+

.= ro

,

z

.

,

I

.

;

:

:

:

:

O1.,...

0

:::::

4

.

o-

* 012345012345

’

s

Horizontal Position (cm)

Fig. 6. Scatter plot of distribution of hits measured in WCl, corresponding to 66.0 MeV protons stopping in AEl, AE2, , AES, respectively (left column). These are compared with Monte Carlo simulations of similar statistics (right column).

Fig. 5. Phantom used to compare with simulations. Its position relative to the proton beam is shown in Fig. 3.

data and the simulation. On the top of Fig. 6 we sketch the region of the phantom spanned by the beam. A more detailed comparison can be made for this particular phantom taking advantage of the symmetry along the Y-axis (Fig. 6), and Fig. 7 shows the corresponding projection along the X-axis. Within the statistical uncertainty, excellent agreement is observed between the data



and the simulation. Note the change in scale for AEl and AE5, and the decrease in counts for the second structure at AE4 in relation to the nearby peaks. Fig. 8 shows the results of scanning the region around this minimum (between 1 and 3 cm, see Fig. 7) with better statistics. Again, the spectra in detector AE4 shows a marked decrease in counts for the second peak, as predicted by the simulation: the relative peak intensity in AE3 is 140/120 = 1.2, while

563

Dam

slmhti

1

Data

am

160

100

D

0 60

100

50

60

t 0

100

60 50

0

6

.

0

1

2

3

4

5

0

1

2

3

4

100

-

50

-

1

5

2

1

2

3

Position (cm)

Position (cm) Fig. 7. Projection on the horizontal (X-) axis corresponding to the scatter plots of Fig. 6.

3

Fig. 8. As in Fig. 7, but scanning only the region l-3 cm three times

the statistics.

with

J.I.. Romero et al. / Nucl. fnstr. and Meth. in Phys. Res. A 356 (19951558-565

564

= 4.7. This corresponds to a thickness difference of 0.5 mm, or J 0.05 g/cm’ (compared to an overall thickness of 30 mm). The simulations predict this dip quite well. For the purposes of evaluating the quality of the Monte Carlo calculations, we believe that the method of presentation provides the most detailed comparison. We did not attempt to do tomographic reconstruction of the phantoms. That important aspect for radiotherapy is beyond the scope of this paper.

Simulation

in AE4 is 140/30

5. Simulations

r-----l

at higher energies

The good agreement observed between the data and the simulations for 66 MeV protons gives us confidence in our Monte Carlo simulations and allows us to make predictions on the quality of the method for higher energy protons. An example of such predictions is presented in Fig. 9. The setup corresponds to that of Fig. 2. The phantom. shown on the left on Fig. 9, is (for simplicity) a

Simulation 2 ~E16

0

2

4 Horizontal

6

13 Postion

10 (cm)

Fig. 10. Simulation of phantom at 245 MeV proton energy. The top panel plots the detector number where the proton stops versus the horizontal position, averaging over 100 hits per bin. In order to see the intensity, the hits have been randomly spread by 0.8 units around each integral vaiue corresponding to each detector (vertical axis). The bottom panel was obtained by averaging only 10 hits, that is, by using a factor of 10 fewer protons in the simulation.

2 ~E15

0 2

AE14

0

bE13

AEl2 2 AElI

0 2

AElO

0

0-

0

2

4

6

8

10

Horizontal PosRlon (cm)

Fig. 9. Simulation of phantom at 245 MeV proton energy. The setup for this simulation corresponds to that of Fig. 2. The phantom used is shown on the left, and it corresponds to a cylindrical bone surrounded by 35 cm of water. The proton beam is perpendicular to the axis of the bone. The right figure shows, upper panels, the X-Y distribution of proton hits stopping at the indicated detectors. The bottom panel shows a profile of the horizontal position (see Fig. 10 for details).

cylinder, 2 cm in diameter and 3 cm in depth, and positioned at the center of a surrounding media consisting of Ha0 (35 cm in depth, p = 1.0 g/cm3). We assume the cylinder to have a composition similar to that of bone, i.e., CaO, ( p = 1.5 g/cm3). The proton beam, with an energy of 245 MeV, impinges perpendicular to the main axis of the cylinder (Fig. 9). We send 3 X lo5 protons over an area of 7 cm (horizontal) by 3 cm (vertical). The X-Y positions correspond to that of WC1 (u = 1.3 mm). The data is binned in X-Y areas of 1 mm2, and every 100 hits are averaged in each bin, to take advantage of statistical sampling. The top part of Fig. 9 (right) shows the average (100 hits per bin) X-Y distribution at WC1 corresponding to protons stopping in detectors AElO through AE16. The cylindrical bone was positioned with its center at X = 6 cm. The bottom panel of Fig. 9 (right) plots the detector where each proton stops (average over 100) versus the X position, and provides another way of displaying the results. (This is possible because of the symmetry of this phantom along the Y-axis.) The bone structure is discerned quite well even with much less statistics, as seen in Fig. 10, where we compare the results with the case corresponding to averaging only every ten hits (i.e., ten times less integrated beam intensity, or 3 X lo4 protons). As mentioned above, a more involved tomographic reconstruction has not been pursued, since we have dealt with

J.L. Romero et al. /Nuct.


rather simple phantoms in our tests and simulations. Future work should involve more complex phantoms and possibly tomographic reconstruction. It should be pointed out that the beam time involved, as well as the dose, necessary to obtain a radiography to position a patient using this method are very low [lo]. As an example, consider a very low beam current of only 5 X lop5 nA, which corresponds to 3 X lo5 protons/s. This rate is entirely within the counting rate capabilities of the detectors and fast electronics. To obtain a radiography like that in Fig. 4, but at 245 MeV, one would need about 3 X 10” protons, which will take 10 s of irradiation time on the patient. Data processing time should be of the order of seconds.

6. Conclusions We have described a method to locate density variations in phantoms that makes use of high energy proton beams. The method is thought to be applicable to the positioning of a patient in a proton therapy facility. The method makes use of a proton range telescope for density variation and X-Y position sensitive detectors for planar positioning. We have tested the principle of the method with measurements at a proton energy of 66 MeV. Good visual quality is seen in the tests. We have compared the measurements with detailed Monte Carlo simulations and achieved good agreement. Further simulations with high energy proton beams (245 MeV) show that the method should provide good visual quality and sensitivity for positioning. In the case of a proton therapy facility, the latter could be achieved by using the proton therapy beam per se in a (very) low intensity mode.

Acknowledgement One of the authors (H.K.) was partially supported by UCSF REAC Simon grant. We are pleased to acknowledge the support of the National Science Foundation Grant

565

PHY91-23301. The support and willing assistance of the staff of Cracker Nuclear Laboratory is gratefully appreciated.

References [ll G. Coutrakon, M. Bauman, D. Lesyna, D. Miller, J. Nusbaum, J. Slater, .I. Johanning, J. Miranda, P.M. DeLuca Jr., J. Siebers and B. Ludewigt, Med. Phys. 18 (1991) 1093. [z] K. Kawachi, T. Kanai, M. Endo, Y. Hirao and H. Tsunemoto, J. Jpn. Ther. Radiol. Oncol 1 (1989) 19. [3] K.P. Gall, L. Verhey, J. Alonso, J. Castro, J.M. Collier, W. Chu, I. Daftari, G. Goitein, H. Kubo, B. Ludewigt, J. Munzenrider, P. Petti, T. Renner, S. Rosenthal, A. Smith, J. Staples, H. Suit and A. Thornton, Nucl. Instr. and Meth. B 79 (1993) 881. [41 L. Verhey, M. Goitein, P. McNulty, J. Munzenrider and H. Suit, Int. J. Radiat, Oncol. Biol. Phys. 8 (1982) 289. [51 I. Daftari, P.L. Petti, J.M. Collier, J.R. Castro and S. Pitluck, Med. Phys. 18 (1991) 1105. [61 M.L. Graham, A.Y. Cheng and L.Y. Wong, Int. I. Rad. Oncol. Biol. Phys. 20 (1991) 613. [71 A.M. Koehler, Science 160 (1968) 303. P31V.W. Steward and A.M. Koehler, Surg. Neurology 2 (1974) 283. 191 V.W. Steward and A.M. Koehler, Radiology 110 (1974) 217. [lOI K.M. Hanson, IEEE Trans. on Nucl. Sci. NS-26 (1979) 1635. ill1 K.M. Hanson et al.. IEEE Trans. Nucl. Sci. NS-25 (1978) 657. WI K.M. Crowe, T.F. Budinger, J.L. Cahoon, V.P. Elischer, R.H. Huesman and L.L. Kanstein, IEEE Trans. Nucl. Sci. NS-22 (1975) 1752. b31 K. Kouris, N.M. Spyrou and D.F. Jackson, Imaging with ionizing radiations (Surrey Univ. Press, 1982). [141 C.A. Tobias et al., NC1 401 CB 40302 (1976). [151 H. Bichsel, American Institute of Physics Handbook, Coordinating Ed. D.E. Gray, (McGraw-Hill, 1972) chap. 8d. D61 J.L. Romero, F.P. Brady, G.A. Needham, C.M. Castaneda, T.S. Subramaniam, J.L. Ullmann and C.I. Zanelli, Nucl. Instr. and Meth. 171 (1980) 609. 1171 T. Ford, G.A. Needham, F.P. Brady, J.L. Romero and C.M. Castaneda, Nucl. Instr. and Meth. 228 (1984) 81. iI81 J.L. Romero, unpublished (1994).