GPU-accelerated bi-objective treatment planning for prostate high-dose-rate brachytherapy

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Brachytherapy treatment planning involves complex optimization to balance target coverage and organ sparing.
• Previous bi-objective methods for prostate HDR brachytherapy offered trade-offs but were computationally intensive.
• Clinical acceptance requires efficient planning methods that deliver high-quality treatment options.

Purpose of the Study:

• To enhance a bi-objective treatment planning method for prostate HDR brachytherapy.
• To reduce computational runtime while improving treatment plan quality.
• To achieve clinically acceptable execution times for advanced treatment planning.

Main Methods:

• Utilized a state-of-the-art multiobjective evolutionary algorithm for bi-objective treatment planning.
• Implemented graphics processing unit (GPU) acceleration for parallel dose calculations and DV index computation.
• Modified optimization objectives to improve the quality of generated treatment plans.

Main Results:

• Achieved optimization results in 30 seconds, a significant reduction from previous 1-hour runtimes.
• Generated hundreds of potential treatment plans per patient, offering diverse trade-offs.
• Successfully identified plans meeting clinical protocol criteria for 15 out of 18 patients, compared to 4 out of 18 previously.
• Increased DV index calculation accuracy (100,000 points) within 3 minutes yielded higher quality plans.

Conclusions:

• GPU-accelerated bi-objective planning for prostate HDR brachytherapy now yields large sets of high-quality plans rapidly.
• Optimization within 3 minutes, even with increased accuracy, is clinically feasible and enhances plan selection.
• The improved method facilitates more informed clinical decisions by providing a wider range of optimized treatment plans.