Encrypted login | home

Program Information

A Method to Accelerate Monte Carlo Calculation Based QA for MRI Guided Online Adaptive Radiation Therapy

Y Wang

Y Wang*, T Mazur , J Park , D Yang , S Mutic , H Li , Washington University School of Medicine, St. Louis, MO


TU-D-205-1 (Tuesday, August 1, 2017) 11:00 AM - 12:15 PM Room: 205

Purpose: MRI-guided online adaptive radiation therapy enables plan re-optimization based on on-board imaging immediately before treatment delivery. The purpose of this work is to develop a method to accelerate Monte Carlo calculation-based patient specific QA.

Methods: The patient’s online MRI is divided into two parts: important region (IR) where clinical relevant DVHs are generated and the rest unimportant region (UR). Detailed Monte Carlo simulation are performed in all IR voxels. If one voxel is labeled as UR, the distance to the nearest IR will be pre-calculated and assigned to it. When simulating a secondary electron transporting in an UR voxel, we first estimate its CSDA range in current material. If it cannot reach the IR, we simply stop the simulation and deposit its energy into current voxel. Otherwise, detailed electron transport simulation is followed. The brutal termination is reasonable due to two facts associated with the ViewRay’s MRIdian platform: (1) the secondary electrons generated from a Cobalt photon can only travel one or two voxels (3 mm voxel) prior to a complete stop; (2) the electron returning effect caused by the magnetic field shortens the distance that an electron can travel. We additionally force all low-density voxels inside the body to be labeled as IR.

Results: We implemented the above algorithm and performed benchmarks based on our latest online adaptive treatment plan Monte Carlo verification toolkit, which incorporated both GPU acceleration and variance reduction into the widely accepted DPM package. The simulation time for 10 million history reduces by 50% to about 1 minute using the proposed algorithm, producing the same PTV-DVH.

Conclusion: The proposed method can finish Monte Carlo based online IMRT QA in about 1 minute.

Contact Email: