iPlan MC dose calculation features

The following four options can be changed by the user in the MC dose display settings window:

• spatial resolution (in mm),

• dose result type (dose-to-water or dose-to-medium),

• mean variance (in mm), and

• MLC model (Precise or Fast).

The spatial resolution parameter defines the size of the MC dose computation grid. The default resolution size is 5 mm. For small target volumes it is recommended to use smaller grid sizes (2-3 mm).

The mean variance parameter is related to the statistical uncertainty of the MC calculation. The mean variance estimates the number of histories required to achieve this variance per beam in percentage of the maximum dose of that beam (normalized to that beam). Therefore, the variance in overlapping regions is less than the specified mean variance. The default setting is 2%.

water (Dw). The MC dose calculation does inherently compute (Dm) - energy de-

posited per unit mass in the tissue of interest. The conversion to (Dw) is performed

by invoking Bragg-Gray cavity theory:

Dw = Dm  S ρ w m (4.2) , where  S ρ w

m is the unrestricted water-to-medium electron mass collision stopping

power averaged over the energy spectrum of the primary electrons.

The MLC model can be set to Precise or Fast modes. In the first mode, all the features of the MLC geometry, such as the gap between leaves, rounded tip leaves, leaf thickness, etc., are taken into account whereas with the Fast mode the calculation times decrease, but does not take into account some details of the MLC specifications.

Pencil-beam and collapsed-cone algorithms are different implementations of the superposition/convolution technique. They are based on several approximations and simplifications such as:

• the linac head is modeled by simple sources, i.e., point sources or parallel sources;

• the electron transport is modeled on straight lines;

• neglect or simplify modeling the lateral density scaling (slab approximation), in the case of pencil-beam algorithms;

• discretization of the point-spread function (energy kernel), in the case of collapsed-cone algorithms;

• neglect or approximate representation of the energy kernel tilt angle to reduce computation time;

• neglect or estimation of effects such as depth hardening and off-axis softening.

Usually these techniques provide results within seconds but the dose calculation can be incorrect especially in the head and neck and thorax regions due to the approximations used.

Several groups have already developed MC models in independent MC codes to verify TPS algorithms, specially for more complex techniques such as IMRT [135].

There are evidences pointing to PBC overestimating the doses in air regions such as lung and head and neck localizations, up to 10% when compared to MC calcu- lations [181].

Other studies involving verification of TPS (convolution/superposition algorithm: Collapsed Cone approximation algorithm) with MC (BEAM EGS4-based MC code) used monitor units calculation comparison particularly for narrow and ir- regular segments was performed by Francescon et al [182]. Maximum differences of 8% were found for such segments where there is electronic disequilibrium and inhomogeneities which become of paramount importance.

Carrasco et al [183] also performed MC simulations using PENELOPE code and performed comparisons with several TPS correction-based and superposition/- convolution algorithms for percentage depth doses (PDD) measured with TLD, metal-oxide semiconductor field-effect transistors (MOSFETs), plane parallel and cylindrical IC, and beam profiles with films. The superposition convolution algo- rithm had a better performance than the corrected-based algorithms especially in the build up region.

In the case of this study, the calculation algorithms the TPS has, are the Pencil Beam Convolution (PBC) and the commercial Monte Carlo (iMC). The required data was measured and introduced in iPlan BrainLAB TPS and the data was afterwards validated according to the BrainLAB requirements and specifications. This TPS has a license to perform calculation in arc gantry mode which was adopted for Dynamic Conformal Arc RT (DCART). However, it does not allow the use of wedges. The modulation was accomplished by using field in field technique and smaller segments, whenever necessary, to achieve the beam irradiation goals.

Comparison of iPlan algorithms: PBC and iMC

iMC was verified by K¨unzler et al [184] using gamma analysis performed measure- ments and compared the results with the calculations. They concluded that, even in heterogeneous media, the iMC algorithm leads to accurate dosimetric results. Fragoso et al [185] also performed a dosimetric and clinical evaluation of iMC and achieved good agreement in homogeneous and heterogeneous media.

Ali et al [186] performed a comparison between PBC and iMC for several patholo- gies (brain, prostate, lung, head and neck and paraspinal tumors). Dose volume histogram analysis showed that agreement within 5% for all cases except in lung

tumors where discrepancies can be up to 30% with MC not covering the lung tumor against PBC prediction.

Petoukhova et al [187] explored further the heterogeneous pathologies such as lung and head and neck by performing IC and film measurements and achieved excellent agreement in homogeneous media for several static shaped and IMRT fields between iMC and the measurements. In the presence of heterogeneities, iMC accurately predicts the dose and it accurately takes into account the lateral electron transport. The sample of patients chosen had breast sizes varying from 350 to 1750 cm3_{; no correlation between the breast size and the irradiation doses}

was in the scope of this study. Das et al [188], in fact, claim there is a relationship between the breast volume and the CLB doses; however, Zurl et al [189] studies did not find a relationship between these parameters, after analyzing the anatomic and field specific parameters.