Click here to

Session: Therapy General ePoster Viewing [Return to Session]

Implementation of a Static and Dynamic MLC Model in a GPU-Accelerated Monte Carlo Engine

E Brost, H Wan Chan Tseung, J Antolak*, Mayo Clinic, Rochester, MN


PO-GePV-T-264 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: To fine-tune the implementation of the Varian TrueBeam multi-leaf collimator (MLC) geometry in a GPU-accelerated Monte Carlo (MC) engine and include dynamic MLC simulations of electron and photon transport.

Methods: A CUDA framework was previously developed for the calculation of electron fields collimated using the conventional MLC, modeled using specifications provided by the vendor. However, in our initial implementation, discrepancies in the calculated dose were observed in regions collimated using MLC leaves >10 cm from isocenter. A static field with a complex collimation pattern was used as input to the MC to calculate electron dose deposition in a solid water phantom, which was compared to radiochromic film and ion chamber array measurements. The results were used to fine-tune and adjust the MLC geometry. A dynamic MLC model was implemented. To simulate continuous leaf motion, a random jaw/MLC position is sampled independently for each primary particle, and the MLC geometry is rapidly initialized using plane and curved surfaces parameterized by simple equations.

Results: Adjusting the MLC geometry based upon their projected width at isocenter, while preserving proper divergence, tongue-and-groove and vendor-provided geometric ratios, improved the agreement between measured and calculated dose distributions. Gamma passing rates after MLC adjustment were 99.6%, 99.1%, and 100.0% for 6, 12, and 20 MeV film measurements using a 3%/3mm criteria.

Conclusion: Following the vendor specifications alone led to discrepancies in dose calculations at the edge of the field. Improvements were made by considering the projection width of MLCs to isocenter, while preserving the vendor-specific ratios of MLC leaf dimensions. The MLC model implemented in the code allows for complete sampling of dynamic MLC sequences.


Monte Carlo, Electron Therapy, MLC


TH- External Beam- Photons: Computational dosimetry engines- Monte Carlo

Contact Email