Exploration of Reduced Order Prioritized Optimization for IMRT Treatment Planning
G Kalantzis1*, L Rivera2, A Apte3, R Radke4, A Jackson5, (1) Memorial Sloan-Kettering Cancer Ctr, New York, NY,(2)Rensselaer Polytechnic Institute, Troy, NY,(3)Memorial Sloan-Kettering Cancer Ctr, New York, NY,(4) Rensselaer Polytechnic Institute, Troy, NY, (5) Memorial Sloan-Kettering Cancer Ctr, New York, NYSU-E-T-600 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Investigate the feasibility of reduced order prioritized optimization for IMRT treatments.
Methods:The proposed method consists of three stages. Firstly, we sample the intensities space by solving a series of unconstrained optimizations of a scalar weighted sum of partial objectives for the target and the organs at risk (OARs). Secondly, the dimensionality of the search space is reduced using principal component analysis on the solution samples. Finally, treatment planning goals/objectives are prioritized and the problem is solved sequentially in the reduced order space: low priority objectives are optimized provided they do not interfere with the higher priority objectives. In the current study a quadratic deviation of the prescribed dose and the mean dose was used for the objectives of the PTV and the OARs respectively. Finally, a slip factor s, a dimensionless parameter, was used in order to relax the hard constraints regarding the PTV coverage within the established max and min dose limits and offer more flexibility to the algorithm for the lower order priorities.
Results:The method was applied to a prostate IMRT plan with five coplanar beams. Two OARs were considered: rectum and bladder. On completion of the sequential prioritized optimization the mean dose at the rectum and the bladder was reduced by 20.5% and 21.5% respectively compared to the mean dose at the first step of the optimization for a slip factor s = 3. Finally, a max speedup of ~22 was achieved for the prioritized optimization steps.
Conclusion:A proof of concept was demonstrated for reduced order prioritized optimization for IMRT planning. Dimensionality reduction techniques may be applied to reduce the complexity time of the constraint prioritization optimization steps for IMRT planning.