For the commissioning of the linac, the MC simulations performed in DOSXYZnrc, were done in a homogeneous water voxelized phantom.
For the patient simulations, a 3D voxelized phantom converted from the patien- t/phantom CT scan was used for dose calculation. The density and material for each voxel were converted based on the CT number.
3.8.1 CTCreate
A phantom is constructed using a 3-D matrix of voxels (volume elements) for which each voxel contains a physical density and material assignment. There are two ways by which this matrix is most commonly created. The first way is by simply defining a set of x, y, and z boundaries and assigning densities and material types to these voxels using an (.egsinp) input file. The second way, used commonly for simulations on patient geometries, is to create the phantom from a set of CT images taken of the patient. From this set of images, the CT densities in Hounsfield units (HU) are interpolated onto a 3-D matrix of voxels and converted into equivalent physical densities
Typically, the CT-numbers range between -1000 HU (air) to +1000 HU (bone). The conversion function for any individual CT scanner is unique and obtained based on the CT calibration with a standard phantom. The material type (com- position) and mass density data within each voxel are derived from the Hounsfield
number exported from the CT using a CT conversion curve, included in table 3.2. More details on the HU and calculation considerations can be found in 4.2.2.1.
Table 3.2: CT numbers and density range for the four materials used in the ramp for converting CT numbers to material parameters (composition and density).
Air Lung Tissue Bone
CT number range 0 - 50 50 - 300 300 - 1125 1125 - 3000 Density range 0.001 - 0.044 0.044 - 0.302 0.302 - 1.101 1.101 - 2.088
The conversion from CT densities to physical densities is achieved through inter- polation of optical-to-physical density relations forming a CT ramp. The matrix of physical densities is then written to a file (.egsphant). Also included in this file is a list of media present in the patient model, as well as a map of individual voxels to material type. It should be noted that, in general, not every material type is defined in this file. Instead, only the most common material types may be included.
The BEAMnrc software package includes CTCreate, a tool for creating phantom models, as part of the distribution. The DICOM (Digital Image and Communi- cations in Medicine) image format provides a standard format by which the CT scanner data can be exported. The image resolution is not fixed, although 512 × 512 pixels is commonly used. The spacing of the pixels is determined by the size of the area that was imaged and is uniform along each axis. The number of CT images or slices is generally determined by the volume of the prescribed region to be imaged and the slice spacing specified by the oncologist.
3.8.2 DOSXYZnrc
DOSXYZnrc [122] is another user code for the EGSnrc system dedicated to the calculation of dose distributions within phantoms that can be defined either by voxels or extracted from CT data.
Once particle transport through the accelerator head has been simulated with BEAMnrc, the output can be transported into the constructed patient phantom using the DOSXYZnrc user code (included in the BEAMnrc distribution). There are several ways in which the particles emerging from the accelerator head may be passed on to DOSXYZnrc. First, through the use of a phase space file generated in BEAMnrc. Second, by characterizing the particles from BEAMnrc into a series of histograms and sampling from the derived particle source model. Lastly, by incorporating the BEAMnrc simulation as a shared library in DOSXYZnrc such
that each particle requested by DOSXYZnrc is transported through the BEAMnrc simulation on-the-fly.
DOSXYZnrc is specifically written for obtaining the dose (and accompanying un- certainty) in a cartesian geometry. In DOSXYZnrc, a spherical coordinate system is defined to describe the beam incidence on the phantom with an origin set to the isocenter, as defined by (xiso,yiso,ziso) the x, y, and z distances from the de-
fined (0, 0, 0) origin of the phantom. The incident beam angle is specified by θ, φ, and φcol (see figure 2.7 from DOSXYZnrc manual). In practical use of the
accelerator, the incident beam angle is more commonly defined by specifying a gantry angle, couch rotation angle, and collimator rotation. A coordinate trans- formation is then required. The absorbed dose is recorded or scored in a 3-D array of voxels with boundaries as defined in the CT phantom. The user can choose to omit specific voxels from the scoring array. By default, the dose, uncertainties, and voxel boundaries are all written in American Standard Code for Information Interchange (ASCII) format to a (.3ddose) file.
When PSF are used as input of the DOSXYZnrc, there is a strategy to increase the number of particles simulated in a run that consists in re-using the PSF. The number of times the PSF is used/recycled is defined through the NRCYCL input parameter. Along with this particle recycling strategy, ISMOOTH is used to redis- tribute the recycled particles about the central axis of the linac beam. Recycling can, nonetheless have a significant effect on the final statistical uncertainty of the dose , possibly creating correlations between particles in the PSF. This effect is much more considerable for electron than for photon beams [95]. In this work, a maximum of 20 recycling factors was used in the dose calculations except for the closed fields, for the adjustment of the MLC parameters.
The results of MC simulations are not immediately comparable to absolute dose calculations and in the commissioning process, relative (normalized) curves are generally used. However, for the calculation of dose distributions, absolute dose is important and depends on the MU necessary to administer the patient in pre- determined conditions. In the case of the linac used, 1 UM corresponds to 1 cGy at an SSD of 100 cm, at 10 cm in depth, for a 10 x 10 cm2 field. It is not possible
to simulate all the particles that are present in a single MU (6.23 × 1010 keV per
cm3, of water). The number of source particles are restricted to the smallest value
capable of producing the desired statistical accuracy, in order to minimize the calculation time. A calibration process is required to obtain the desired absolute dose values.