Program Information
Feasibility of a Fast Inverse Direct Aperture Optimization Algorithm Using Matrix Inversion
M MacFarlane*, D Hoover , E Wong , J Battista , J Chen , University of Western Ontario, London, ON
Presentations
WE-RAM3-GePD-T-4 (Wednesday, August 2, 2017) 10:30 AM - 11:00 AM Room: Therapy ePoster Lounge
Purpose: To develop a fast direct aperture optimization algorithm for optimizing intensity-modulated radiation therapy (IMRT) beams using matrix inversion.
Methods: The algorithm developed is an extension of an exceptionally fast fluence map optimization algorithm called Fast Inverse Dose Optimization (FIDO). FIDO’s objective function was reformulated to optimize a plan’s segment weights and multileaf collimator (MLC) positions, via a matrix equation, without yielding non-deliverable solutions. This fast inverse direct aperture optimization (FIDAO) algorithm has been implemented in matRad (an open source treatment planning toolkit in MATLAB) for feasibility testing. The algorithm was applied to a prostate, a liver, and a complex head-and-neck cancer case and compared with matRad’s built-in direct aperture optimization algorithm which uses an interior point optimization (IPOPT) algorithm to optimize a conventional aperture-based objective function. To minimize planning bias, the same IMRT objectives and initial MLC segments were used for both FIDAO and IPOPT algorithms. Optimization times and plan quality were compared between FIDAO and IPOPT optimized plans.
Results: For every test case, the new FIDAO algorithm provided superior tumor volume coverage with significantly less dose to nearly all surrounding organs at risk and shorter optimization times. Optimization times for the prostate, liver and head-and-neck cases were 0.8s vs. 58.8s, 0.3s vs. 25.2s, and 15.4s vs 25.8s for FIDAO and IPOPT, respectively. The suboptimal plan with IPOPT in the head-and-neck case might be a result of the algorithm becoming trapped at a local minimum. The increased FIDAO optimization time compared to other cases was a result of the larger number of iterations needed for the head-and-neck case, as well as the larger number of optimization parameters due to the increased field size.
Conclusion: We have demonstrated the feasibility of a fast inverse direct aperture optimization algorithm with promising dosimetric results and speed in a few IMRT test cases.
Funding Support, Disclosures, and Conflict of Interest: This work was funded by a tri-partisan research agreement between the London Health Research Institute, Philips Healthcare, and the Government of Ontario.
Contact Email: