One of the most important issues limiting the use of Monte Carlo techniques for clinical
absorbed dose calculations is the long execution time required to reduce the statistical variance
associated with the calculated absorbed dose at each differential volume. At the same time, the
outcome of any radiotherapy procedure depends solely on the accuracy of the radiotherapy
treatment planning system, so that a small change in the absorbed dose delivered (A 5%) can
result in a dramatic change in local response of the tissue (f 20%)12,13
In order to reduce the statistical variance a, a ac associated with Monte Carlo, large
numbers of histories N are required, and consequently this leads to higher running time to yield
acceptable results. Improved variance reduction techniques may reduce the running time.
However, such techniques are to be used with caution, as improper application of statistical
variance reduction, method applied un-conservatively may cause wrong answers. Also, variance
reduction is often tied to one objective; this can be problematic when faced with determining a
dose at all locations in the phase space.
The increasing use of voxelized human body geometries has created challenges in many
aspects of applying Monte Carlo simulationS14. With increasing use of modern technologies such
as CT and MRI in vivo imaging, anatomical definition has shifted from using simplified surface
equations to segmented tomographic images in order to truly reflect the anatomy of a specific
human body. The segmented images are represented in the form of a large uniform 3-D matrix of
hexahedral voxels, where each voxel represents a constant material with a tag number (for
example, CT number). One digital image may contain billions of voxels, making it very
challenging be fully represented explicitly using Monte Carlo, particularly considering large
models size and number of tally sites, adding significant overhead to the computation. Some
special data management treatments are needed for Monte Carlo simulations to make