At our facility we have Eclipse treatment planning software.
After speaking to our physicists,
they explained that we use an Analytical Anisotropic Algorithm (AAA) at our clinic for photon beams. This is a model-based algorithm. When our machine was commissioned our physicists took measurements of beam output at various depths and fields sizes, to obtain beam profile and symmetry information as well as necessary information such as the percent depth dose (PDD) for all field sizes and energies. This information is used in each calculation to determine monitor units. One drawback to this algorithm is that it looks at heterogeneity as “partial water” for example 4 cm of .25 density lung would be viewed as 1 cm of water. This causes a dose variation of around 3-5% at air tissue interfaces within the body compared to Monte Carlo algorithms; however, the AAA is very comparable in all other aspects and is much faster at performing calculations, which makes it more practical for use in the clinic. For electron beams, we use Monte Carlo. This algorithm simulates how all particles interact from the time of origination to reaching the patient. This results in a more accurate dose calculation because it takes into account dose as a result of interactions with the jaws, detectors, the mirror, mylar, multi leaf collimators (MLC’s), etc. The probability of these interactions happening as a result of the beam energy and field size chosen is also figured into this calculation. With the Monte Carlo algorithm, all density is seen as it truly is, not as “partial water” as in the AAA. This true depiction of density provides for a more accurate dose calculation at air tissue interfaces, making this algorithm superior in this aspect. However, one drawback to the Monte Carlo algorithm is the amount of time it takes to calculate beam data due to the many different factors it takes into account. One question I have when discussing this topic is how, with the advances in technology and computing speed that have been made, does it still take so long to calculate beam data for the Monte Carlo algorithm?