A derivative-free multistart framework for an automated noncoplanar beam angle optimization in IMRT

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Optimization

Background:

• Intensity-modulated radiation therapy (IMRT) planning involves selecting radiation beam angles.
• Beam angle optimization (BAO) can improve IMRT plan quality but is rarely used clinically.
• Noncoplanar beam incidences offer potential advantages for specific tumor sites but are complex to optimize.

Purpose of the Study:

• To present the benefits of a derivative-free multistart framework for optimizing noncoplanar beam angle selection in IMRT.
• To evaluate the effectiveness of this framework in improving treatment plan quality for nasopharyngeal cancer.

Main Methods:

• A derivative-free multistart optimization framework was developed to explore the noncoplanar beam angle optimization problem.
• The framework combines global search with local optimization to avoid entrapment in local optima.
• Twenty nasopharyngeal cancer cases were analyzed, comparing plans generated by the framework against coplanar and equispaced beam setups.

Main Results:

• The optimized noncoplanar beam setups resulted in superior treatment plans compared to equispaced or optimal coplanar setups.
• Significant improvements were observed in target coverage and organ sparing.
• Specific reductions included an average of 6.1 Gy in oral cavity mean dose and 7 Gy in brainstem maximum dose compared to equispaced plans.

Conclusions:

• The noncoplanar beam angle optimization problem is a challenging multimodal optimization task.
• The proposed derivative-free multistart framework effectively addresses this challenge, yielding high-quality IMRT treatment plans.
• This framework offers a promising approach for automated noncoplanar beam selection in modern IMRT planning.