Fifty-eighth annual meeting of the american association of physicists in medicine
SU-D-BRB-06: Toward Big-Data Driven Treatment Planning: VMAT/IMRT Inverse Planning Autopiloted by Population-Based Prior Data
Current treatment planning remains a costly and labor intensive procedure and requires multiple trial-and-error adjustments of system parameters such as weighting factors and prescriptions. The purpose of this work is to develop an autonomous treatment planning technique with effective use of population-based prior knowledge.
Our autopiloted planning tool consists of three major components: (i) a commercial treatment planning system (TPS) (EclipseTM, Varian Medical Systems, CA); (ii) a statistical formulation of prior knowledge, which takes an assemble of prior treatment plans similar to the patient under planning and provides a priori variation range of final solution with an assigned preference level; and (iii) a decision-making or outer-loop optimization that assesses plan generated by Eclipse and drives the search toward a solution consistent with population-based prior knowledge. To query and interact with TPS, Microsoft (MS) Visual Studio Coded UI is applied to record some common planner-TPS interactions as subroutines. These subroutines are called back in the autopilot program written in C# to navigate through the solution space with an iterative algorithm. The utility of the approach is demonstrated by using a prostate IMRT case and three head and neck VMAT case.
An autonomous inverse planning formalism of using population-based prior knowledge for autopiloting the VMAT/IMRT treatment planning process is implemented successfully in platform of a commercial TPS. The process mimics the decision-making process of autopiloted automobile and provides a clinically sensible treatment plan automatically, thus effectively eliminating the tedious manual trial-and-errors process of treatment planning. It is found that the prostate and head and neck treatment plans compare favorably with that used for patients’ actual treatments.
Clinical inverse treatment planning process can be autopiloted effectively with the use of population-based prior knowledge. The strategy lays foundation for future development of big data-driven inverse planning of various disease sites.
Research grant and speak honoraria from Varian Medical Systems