Program Information
Threshold-Driven Optimization for Reference-Based Auto-Planning
T Long*, M Chen , S Jiang , W Lu , UT Southwestern Medical Center, Dallas, TX
Presentations
SU-C-BRB-7 (Sunday, July 31, 2016) 1:00 PM - 1:55 PM Room: Ballroom B
Purpose: To study the procedure of reference-based auto-planning for treatment plan optimization and develop a threshold-driven optimization methodology for automatically generating a treatment plan that is motivated by a reference DVH and/or dose distribution. We study threshold-driven optimization for reference-based auto-planning (TORA).
Methods: Commonly used voxel-based quadratic penalties have three components: an overdose weight, and underdose weight, and some target dose threshold. Conventional planning using such a function involves iteratively updating the preference weights while keeping the thresholds constant, an unintuitive and inconsistent method for planning towards some target DVH. In auto-planning, current cutting-edge techniques also focus on the preference weights while fixing the threshold values. However, driving the optimization by threshold values can achieve similar plans without the extreme updates usually required of the weight parameters. The proposed methodology directly relates reference information to threshold values, which influence the optimization in a more impactful, intuitive way than the preference weights.
Results: This methodology was applied to a prostate cancer case as a proof-of-concept. Both IMRT and VMAT treatment planning models were able to closely achieve a reference DVH. Other cases and modalities will be studied.
Conclusion: This auto-planning technique has the potential to be very effective for auto-planning based on reference plans. As dose prediction and data-driven planning becomes more prevalent in the clinical setting, incorporating such data into the treatment planning model in a clear, effective way will be crucial for efficient planning. Conventional methods of determining objective function parameters for a voxel-based quadratic objective function have shortcomings and a threshold-focused objective tuning should be explored, especially in the realm of efficient auto-planning.
Funding Support, Disclosures, and Conflict of Interest: Supported by Cancer Prevention & Research Institute of Texas (CPRIT) - ID RP150485
Contact Email: