The complexity and ever-changing nature of the radiation fields generated makes it particularly important to verify treatment plan doses. Researchers from the University of Victoria, and the British Columbia Cancer Agency's Vancouver Island Centre, Canada, have now demonstrated a way of doing this with Monte Carlo methods (Phys. Med. Biol. 53 N359).
This latest work builds on the Vancouver Island Monte Carlo (VIMC) system, a process for verifying radiotherapy treatment plans that can be exported in DICOM-compatible format (Radiother. Oncol. 84 S49). University of Victoria graduate student Karl Bush and colleagues have now incorporated features specific to RapidArc to produce VIMC-Arc.
The result, according to the authors, is: "a fully automated system that constructs Monte Carlo and beam and patient models from a standard RapidArc DICOM dataset, simulates radiation transport, collects the resulting dose, and converts the dose into DICOM format for import back into the treatment-planning system."
The DICOM files used by VIMC-Arc represent each RapidArc treatment as a series of fields with 177 gantry positions, each associated with the relevant multileaf collimeter field aperture. VIMC-Arc models the gantry arc rotation as a series of static gantry positions. Mean gantry angles are computed from pairs of "control points" that describe the position of each linac component.
Intensity-modulated radiation therapy (IMRT) collimator motion is then created from leaf positions at neighbouring mean gantry angles. This step implies that each leaf moves with constant speed as the gantry rotates between two control points, though the speed of individual leaves may vary. The result is 176 sliding window IMRT radiation "fields".
According to Bush and colleagues, previous Monte Carlo verification models for intensity-modulated arc therapy (IMAT) involve approximations that make them unsuitable for use with RapidArc. These include using a fairly coarse angular grid to describe gantry positions, modelling dynamic IMAT fields as fixed apertures and ignoring interleaf leakage. Such assumptions are not made with VIMC-Arc.
To test VIMC-Arc, the researchers generated several RapidArc treatment plans on a water-equivalent cylindrical phantom and then recalculated them using their Monte Carlo modelling method. These included a "typical" single-field treatment, a dual-arc treatment, and a RapidArc plan using "avoidance sectors". One single-field RapidArc plan was also calculated from a DICOM patient CT dataset.
Dose distributions produced by VIMC-Arc matched closely with those produced by the anisotropic analytic algorithm (AAA) used in the treatment-planning system. The AAA was known to produce accurate dose distributions in homogenous media. Minor differences between VIMC-Arc and AAA plans were observed in regions with tissue inhomogeneities, as expected. All plans demonstrated better than 1% agreement of dose at the isocentre.
VIMC-Arc is suitable for routine dose verification given that it is fully automated and requires minimal user input, the authors note. Operators may only need to provide the patient's ID, required voxel size and requested dose uncertainty, for example. The system will then do the rest. As such, it is: "an excellent platform for investigations of potential dosimetric problems in different treatment sites with varying degrees of tissue inhomogeneity," they wrote.
The Victoria-based team acknowledge that the accuracy of modelling gantry motion could be improved further by interpolating the 176 segments into several hundred or even thousands. However, they remain confident that this will not be necessary given the anticipated "minute" difference in calculated dose distribution.
Group members are now preparing a quantitative dosimetric analysis of RapidArc dose distributions.