Optimization of intensity-modulated radiation therapy with biological objectives

Area of Science:

• Radiation Oncology
• Medical Physics
• Computational Biology

Background:

• Intensity-Modulated Radiation Therapy (IMRT) treatment planning involves complex nonlinear optimization problems.
• Biological objectives, such as maximizing tumor control probability and minimizing normal tissue complication probability, are crucial for effective IMRT.
• Existing methods often struggle with efficiency and mathematical soundness when dealing with nonlinear objectives and constraints.

Purpose of the Study:

• To develop and evaluate fast, flexible, and mathematically sound variants of the two-metric gradient-projection approach for IMRT treatment planning.
• To address constrained nonlinear optimization problems arising from biological objectives and beamlet weight constraints.
• To present an efficient heuristic for achieving smoother beamlet weight distributions.

Main Methods:

• Formulations with nonlinear objectives based on the Equivalent Uniform Dose (EUD) were considered.
• Fast, flexible variants of the two-metric gradient-projection approach were developed.
• An iterative conjugate-gradient algorithm with an implicit representation of the Hessian matrix was employed for calculating the Newton component of the step.

Main Results:

• The iterative conjugate-gradient approach with an implicit Hessian matrix proved most effective for solving the constrained nonlinear optimization problems.
• An efficient heuristic was developed for obtaining approximate solutions with smoother beamlet weight distributions.
• The proposed methods were validated through testing on a medium-scale clinical case, demonstrating their effectiveness.

Conclusions:

• The developed iterative gradient-projection methods offer an efficient and mathematically sound solution for IMRT treatment planning using biological objectives.
• The implicit Hessian matrix approach combined with conjugate gradients is highly effective for optimizing beamlet weights.
• The heuristic method provides a practical way to achieve improved beamlet weight distributions in clinical settings.