Validation of a rapid algorithm for repeated intensity modulated radiation therapy dose calculations

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Imaging

Background:

• Adaptive radiotherapy and deep learning necessitate rapid dose calculation.
• Existing methods may not meet the speed requirements for these advanced workflows.

Purpose of the Study:

• To evaluate a simple, near real-time algorithm for direct intensity-modulated radiation therapy (IMRT) dose calculation from multi-leaf collimator (MLC) positions.
• To assess the algorithm's speed and accuracy for clinical applications.

Main Methods:

• Developed a novel algorithm using normalized beamlets to modify pre-calculated patient-specific open fields into MLC segment shapes.
• Validated the algorithm on 91 prostate and 20 lung IMRT plans using the Elekta Unity MR-Linac.
• Compared results against Monte Carlo dose calculations using a 3%/2 mm gamma criterion.

Main Results:

• The algorithm achieved high accuracy, matching Monte Carlo dose within 98.02±0.84% for prostate and 96.57±2.41% for lung IMRT plans (3%/2 mm gamma).
• Calculations were rapid: 1.016 ± 0.284 seconds per 9-field IMRT plan for a single patient.
• Parallelized batch processing achieved 0.264 ms per patient, ideal for deep learning.

Conclusions:

• The presented algorithm offers a fast and accurate alternative for IMRT dose calculation.
• It provides a viable option for adaptive radiotherapy and deep learning model training without requiring deep learning model training.
• The method is competitive in speed and accuracy, suitable for repetitive dose calculation needs.