Tissue inhomogeneities in Monte Carlo treatment

planning for proton therapy

L. Beaulieu1, M. Bazalova2,3, C. Furstoss4, F. Verhaegen2,5

(1) Centre Hospitalier Univ de Quebec, Quebec, QC, CA, (2) McGill University, Montreal, QC, CA, (3) Stanford University, Stanford, CA,

(4) Hopital Maisonneuve-Rosemont, Montreal, QC, CA, (5) Maastro Clinic, Maastricht, NL

Introduction

Metal streaking artifacts: The effect of metal streaking artifacts and their

a)

b)

correction based on sinogram interpolation on MC proton beam dose

calculations was studied on a patient with bilateral hip prostheses. Dose

Proton therapy is gaining popularity in the treatment of cancer and the

calculations were performed for three different simulation geometries:

need for an accurate treatment planning system is obvious. Monte Carlo

considering only tissue of uniform density 1 g/cm3 (the water-only

(MC) dose calculation, despite the relatively long computation time, is the

geometry), and using a CT number to material and mass density

most accurate way to determine the dose delivered to the patient during

calibration curve with original CT images containing streaking artifacts

radiation therapy. Whereas MC dose calculations for conventional

and with artifact corrected CT images.

photon and electron radiotherapy have been studied extensively, proton

A treatment plan with two 147 MeV proton beams (45 and 315) was

beam MC dose calculations have only recently received attention. In this

simulated in the MCNPX code. First, the spread-out Bragg peak (SOBP)

work, the importance of tissue segmentation in proton therapy is

was designed using simulations in a uniform water phantom (fig 2a). It is

investigated using dual-energy CT (DECT) imaging. Another challenge in

impossible to model a modulator wheel in the MCNPX code, and

MC dose calculation treatment planning is metal streaking CT artifacts

therefore the steps of the modulator wheel were approximated by 5 mm

with the associated tissue and mass density miss-assignment. Their

thick PMMA blocks. In order to calculate the dose distribution of the

effect on MC proton beam dose calculations is studied for a prostate

SOBP in the patient in a single MC simulation, 11 PMMA blocks were

patient with bilateral hip prostheses.

inserted in the path of the (66) cm2 beams and the source particles (147

MeV protons) were sampled with their respective weights from the

Materials and Methods

volume between the blocks (fig 2b). The patient CT images with

Figure 3: The exact dose distribution (D exact) using a 200 MeV proton beam (a).

PDD with two inhomogeneities (SB3 and B200) (b), the 2% dose calculation

error is indicated by the arrow. The dose differences from D exact for Dsingle (c) and

Ddual (d).

(1.91.920) mm3 voxels were segmented into 4 materials (air, tissue,

bone and steel) using 0.1 g/cm3 mass density bins.

Tissue segmentation with dual-energy CT: CT images of a 30 cm

d)

c)

a)

b)

The dose distribution is significantly distorted in the original CT geometry

due to the artifacts (fig 4b). The apparent air between the prostheses

diameter cylindrical phantom with 9 tissue equivalent inserts (table1, fig

results in inaccurate doses with large statistical errors. Additionally due to

1a) were segmented into material and mass density maps using single-

the air, the 20% and 30% isodose lines extend by 1.5 cm in the healthy

energy CT (fig 1b) and DECT (fig 1c) material extraction. DECT tissue

tissue. This might cause problems in treatment planning and its

segmentation can distinguish materials with similar relative electron

optimization. The artifact corrected geometry produced a dose

densities e having different effective atomic numbers Zeff. The effect of

distribution similar to the water-only dose distribution (fig 4b). The true

inaccurate material segmentation for the two soft bone equivalent

dose distribution is not known.

with the commonly used single energy CT material segmentation was

Figure2: The spread-out Bragg peak for patient dose calculations (a) and the

MCNP geometry showing 0.5 cm PMMA blocks for modulation of the 147 MeV

proton beam (b).

studied. A left lateral 1616 cm2 200 MeV proton beam was simulated in

All CT geometries were converted into lattices and the dose was scored

the MCNPX code. The mass densities for (1.91.920) mm3 voxels were

using the *F8:H,P,E energy deposition tally. Protons, photons and

binned into 0.1 g/cm3 bins. The dose was calculated for the exact

electrons were transported using the la150u cross section library with

geometry (Dexact), the single energy CT geometry (Dsingle) and the dual-

energy cutoffs of 10 keV. In all simulations, 107 particles were simulated

energy CT geometry (Ddual).

in approximately 15 hours on a 3 GHz machine.

materials (B200 and CB2-10) and an adipose-equivalent material (PE)

Table 1:Relative electron densities e and effective atomic numbers Zeff for

materials used in the tissue inhomogeneity study.

Results

MATERIAL

e

Zeff

lung (LN300)

0.292

7.864

lung (LN450)

0.438

7.835

Tissue segmentation with dual-energy CT: Fig 3 presents the results of

polyethylene (PE)

0.945

5.740

the phantom study. The exact dose distribution is shown in fig 4a and the

CT Solid Water (SW)

0.986

8.111

differences from Dsingle and Ddual are presented in fig 3b and 3c,

B200 bone mineral

1.097

10.897

CB2 - 10% CaCO3

1.142

8.905

CB2 - 30% CaCO3

1.286

11.393

CB2 - 50% CaCO3

1.470

12.978

SB3 cortical bone

1.692

14.141

Figure 4: Dose distribution for a

prostate patient calculated on the

basis of homogeneous water

geometry (a), on the basis on the

geometry with metal artifacts (b)

and using the artifact correct

images (c). The arrows indicate the

apparent range of protons due to

artifacts.

b)

c)

respectively. In both Dsingle and Ddual, the position of the Bragg peak is

shifted with respect to the true position of the Bragg peak. The shift is 0.7

cm for Dsingle and 0.7 cm or less for Ddual. This is possibly due to mass

density differences in the single energy CT and DECT geometry from the

Conclusions

exact geometry. Fig 3d demonstrates the dose calculation error in the

miss-assigned B200 soft bone tissue equivalent insert. The dose in the

B200 insert was by 2% lower than in the exact and DECT geometry.

Figure 1: Figure 1: The exact geometry (a), the single-energy material

segmentation (b) and the dual-energy CT material segmentation (c).

a)

The shift in the Bragg peak demonstrates the need for careful mass

density assignment in MC dose calculations for proton beams. The dose

Metal streaking artifacts: The dose distributions calculated based on the

calculation errors using the conventional single-energy CT tissue

water-only geometry, on the original CT geometry and the artifact

segmentation below 2% suggest that the use of DECT for proton dose

corrected geometry are presented in fig 4. The shape of the 80% isodose

calculations might only have a small added benefit. The patient study

line conforms to the prostate in the water-only dose calculation (fig 4a).

shows that a metal artifact correction is necessary for patients with

bilateral hip prostheses.