An ultra-high performance parallel (UHPP) framework for complex 4πradiotherapy planning.

Area of Science:

• Medical Physics and Radiation Oncology
• Computational Science and Engineering

Background:

• Radiotherapy requires precise dose distribution for patient outcomes.
• 4π radiotherapy uses non-coplanar beams for superior dosimetry but faces computational challenges in planning.
• High-dimensional treatment planning is computationally intensive due to large dose-loading matrices.

Purpose of the Study:

• To develop an ultra-high performance parallel (UHPP) framework to accelerate high-dimensional 4π radiotherapy treatment planning.
• To improve computational efficiency without compromising the dosimetric advantages of 4π planning.

Main Methods:

• Developed a two-step TERMA computation module for efficient dose calculation.
• Implemented a synchronized collapsed-cone convolution superposition (CCCS) module with optimized ray sequencing.
• Utilized a scattering-based coordinate transformation for accurate dose mapping.
• Employed a fast iterative shrinkage-thresholding algorithm with group sparsity regularization accelerated on GPUs for beam orientation optimization.

Main Results:

• UHPP achieved high dose accuracy, with minimum 98% gamma passing rates (1.5%/1.5 mm) in phantoms and high average rates in patient cases (e.g., 97.35% for pancreas at 3%/3 mm).
• Delivered significant speedups: 8.86× in dose calculation and 6.99× in plan optimization compared to state-of-the-art methods.
• UHPP and SOTA 4π plans demonstrated superior organ-at-risk sparing and target coverage compared to clinical VMAT plans.

Conclusions:

• The UHPP framework provides substantial computational speedup for 4π radiotherapy planning.
• High dose accuracy is maintained, preserving the dosimetric benefits of 4π techniques.
• This framework supports the practical clinical adoption of advanced 4π radiotherapy.