(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM
Room: Exhibit Hall | Forum 1
Purpose: We design a flexible computational framework that allows one to utilize different optimization schema to design treatment plans. This allows one to evaluate and obtain good results.
Methods: Our modularized framework which consists of 8 components: “Iso-center Determination�, “Beam Selection & Initial Aperture Generation�, “Direct Aperture Optimization�, “Sequencing & Aperture Shape Refinement System�, “Candidate Aperture Shape Pool�, “Aperture Dose Calculation�, “Multi-objective MIP optimization�, and “Plan Evaluation�. A “Feedback System� returns information to the users when no feasible plans can be obtained. In this case, the system returns the most least-violated plans. Using this framework, we generate SBRT treatment plans using a MIP-based beam-angle optimization warm-start, coupled with a gradient-approach MIP-based aperture selection algorithm to generate deliverable optimized apertures.
Results: Fifteen patient cases were tested including five prostate, five lung and five intracranial cases. Our framework is able to generate good feasible plans satisfying all RTOG and clinical requirements. The plans also prove to be superior than the clinical system since some of the clinical plans cannot satisfy all the RTOG requirement. Overall, we observe a 5% reduction in delivery time.
Conclusion: This works indicates the potential for significant dose reduction for organs at risk, DVH improvement and delivery time reduction comparing to current clinical system. Advanced properties (i.e., aperture splitting) have already been implemented in our system and can be inherited. The highly modularized framework can be used to combine different optimization methods to further improve the quality of the final treatment plan. More patient cases and different modules (e.g., column generation based direct aperture optimization can be used in the direct aperture optimization module.) will be investigated.
Keywords
Taxonomy
Not Applicable / None Entered.
Contact Email
