MEDICAL
PHYSICS
UNIT
/
L'UNITÉ
DE
PHYSIQUE
MÉDICALE
McGILL
UNIVERSITY
Monte Carlo (MC) simulations of complex treatment techniques, such as RapidArc, VMAT,
TomoTherapy, tumour tracking, or CyberKnife, require the capability to model continuously
variable beam configurations. The routine QA of these modalities is a challenging task
requiring sophisticated independent methods of treatment delivery verification. For this
purpose, we developed two new sources for the widely used DOSXYZnrc code that support a
continuously rotating gantry and collimator, MLC motion, variable monitor unit rate, couch
rotation and translation in any direction, arbitrary isocentre motion with respect to the patient,
and variable source-to-axis distance. Unique to these sources is the synchronization between
the motion in the patient geometry and the motion of the linac components. The codes that
model the linac and the moving MLC are compiled as shared libraries, which are dynamically
loaded in the computer memory at run time and deliver particles for the patient dose deposition
simulation. Our quasi-continuous, single-run simulations eliminate the need for the hundreds of
intermediate files specific to the traditional brute-force approach and allow an arbitrarily high
degree of spatial and temporal resolution.
We have already incorporated the new sources into our clinical process and routinely use them
as part of our MC based QA for RapidArc. We use ‘dynalog’ files (diagnostic log files
recorded by Varian linacs during the actual beam delivery) as inputs for the MC simulations.
These files contain MLC positions recorded every 50 ms, as well as gantry angles recorded at
specified control points. We use in our MC simulations all of the approximately 2000 different
MLC apertures contained in the dynalog files (as compared to only 177 apertures specified in
the planning files). All RapidArc patient plans verified so far have shown excellent agreement
between MC simulations (typically with 3 billion histories, leading to a statistical uncertainty of
1% in the high dose voxels) and ionization chamber measurements in a cylindrical phantom,
with typical differences around 1%. We also perform a 3D gamma analysis to compare the
Eclipse (AAA algorithm) and MC dose distributions in the full volume of the phantom.
Typically, around 98% of the voxels pass the gamma test, for a 3%/3mm DTA criterion. The
entire QA process is automatic and proves to be highly efficient in terms of departmental
resources and very accurate.