Unencrypted login | home

Program Information

Multi-Criteria Direct Aperture Optimization


E Salari

E Salari*, J Unkelbach, Massachusetts General Hospital, Boston, MA

SU-E-T-656 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: Multi-criteria optimization has been applied to IMRT planning, mainly in the fluence-map optimization stage, to interactively explore the trade-off between conflicting treatment objectives. However, the subsequent leaf-sequencing step may degrade the plan quality due to dose discrepancies. To circumvent this issue, this study aims at developing multi-criteria optimization approaches to IMRT planning using direct aperture optimization methods.

Methods: The direct aperture optimization problem is formulated and solved using a multi-criteria optimization approach. In particular, a solution method is developed to obtain a collection of apertures that can adequately span the entire Pareto surface. To that end, the column generation method for direct aperture optimization is extended to a multi-criteria setting in which apertures that can improve the entire Pareto surface are sequentially identified and added to the aperture collection. More specifically, we approximate the improvement rate of an aperture in the Pareto surface by sampling Pareto-optimal points and measuring the cumulative improvement rate of the aperture at those sampling points.


Results: Our solution method is tested on a paraspinal case in which the trade-off between target coverage and spinal cord sparing is studied. We compare the trade-off curves obtained using our solution method and two other related heuristics. The computational results validate that the proposed method obtains a balanced approximation of the Pareto surface over a wide range of clinically relevant plans.

Conclusion: We have developed a multi-criteria direct aperture optimization method that allows for conducting the trade-off exploration directly in the space of deliverable plans, thereby avoiding dose discrepancies caused by the leaf-sequencing stage.

Contact Email: