Testing of an enhanced leaf model for improved dose calculation in a commercial treatment planning system.

Abstract

With the ever-increasing complexity of dynamic radiotherapy treatments, dose calculation algorithms are challenged to accurately calculate the dose resulting from small, on- and off-axis multileaf collimator (MLC) aperture movements. Although the currently available Eclipse (Varian Medical Systems, Palo Alto) dose calculation algorithms still use a simplified, binary MLC model, a more advanced and detailed modeling of the MLC could be beneficial for the dose calculation precision of high-end treatments. To improve the modeling of the MLC in the dose calculation algorithms of the Eclipse treatment planning system, an enhanced MLC attenuation model was constructed through ray tracing through the actual leaf designs for the most commonly used Varian MLC types. The enhanced leaf model (ELM) thus includes the rounded leaf tip shape, the drive screw cutout, and the leaf body thickness. The purpose of this work is to test out this new model and explore possible improvements compared to the previous model. Dose calculations were performed in a research Eclipse environment equipped with the original and enhanced MLC model. Measurements were performed on TrueBeam and on Halcyon dual MLC treatment units. Dedicated static and dynamic MLC test plans were designed to challenge the dose calculation and highlight differences between both models while keeping the experimental setup simple in order to minimize measurement uncertainties. Measurements were performed with single ion chambers, 2D ion chamber arrays and film. The improved MLC model considerably improves the accuracy of the dose calculation for the test fields used in this study. For the TrueBeam MLC, improvements are most prominent for off-axis dose delivery through narrow (static or dynamic) MLC gaps. For 3mm narrow sweeping gap deliveries at 12cm off-axis, the advanced model agrees within 2% with the measurement, in contrast to the 12% deviation observed with the original MLC model. For the Halcyon MLC, improvements are especially prominent when the leaves of both MLC stacks are aligned, regardless of their position in the field. Sweeping gap measurements improve from a 7%-10% deviation with the original model to within 2% with the new model. Although test fields designed in this study emphasize the flaws in the original MLC dose calculation model, the enhanced MLC model resolves all of the observed discrepancies, showing excellent on- and off-axis agreements with all of the performed measurements.