Penalized likelihood fluence optimization with evolutionary components for intensity modulated radiation therapy...

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Intensity Modulated Radiation Therapy (IMRT) requires precise optimization of beamlet fluences for effective cancer treatment.
• Balancing target dose coverage with sparing of organs at risk (OARs) is a significant challenge in IMRT planning.
• Existing algorithms may converge to suboptimal solutions, necessitating robust optimization strategies.

Purpose of the Study:

• To introduce and evaluate a novel iterative penalized likelihood algorithm with evolutionary components for optimizing IMRT beamlet fluences.
• To assess the algorithm's flexibility in handling various objective functions and dose escalation strategies.
• To determine if optimal beam orientation selection can improve treatment plan quality compared to manually selected orientations.

Main Methods:

• Development of an iterative penalized likelihood algorithm incorporating evolutionary components to avoid local maxima.
• Objective function defined as the product of target equivalent uniform dose (EUD) and homogeneity constraints.
• Application of a quadratic penalty function based on dose-volume histogram (DVH) constraints for OARs.
• Testing on a prostate cancer dataset with challenging DVH constraints for bladder, rectum, and femoral heads.
• Comparison of treatment plans generated with manually selected beam orientations versus globally optimized beam orientations.

Main Results:

• The algorithm successfully generated IMRT plans meeting all DVH constraints for prostate cancer, with average optimization times of approximately 30 seconds.
• Manually selected orientations resulted in homogeneous target dose distributions and adherence to OAR dose limits.
• A global search identified an optimal three-beam orientation (70, 170, 320 degrees) yielding an EUD of 58 Gy, with 96% of the target within limits.
• This orientation-optimized three-beam plan showed comparable or superior target coverage (EUD) to manually selected three-, four-, and five-beam plans.
• The optimized three-beam plan's EUD was slightly lower than the seven-beam plan but demonstrated the potential for fewer beams.

Conclusions:

• The novel penalized likelihood algorithm with evolutionary components is effective for optimizing IMRT beamlet fluences, offering promising results for dose conformity and target uniformity.
• Integrating optimal beam orientation selection with this algorithm can lead to treatment plans with comparable or improved quality using fewer beams.
• This approach holds potential for more efficient and effective prostate cancer treatment planning.
• The algorithm's flexibility and ability to prevent local maxima contribute to robust treatment plan optimization.