Hybrid GPU Monte Carlo software toolkit for fast and accurate cone-beam CT simulation.

Area of Science:

• Medical Physics
• Computational Imaging
• Radiation Oncology

Background:

• Cone-beam CT (CBCT) is increasingly vital for clinical decisions, necessitating faster and more accurate simulation tools.
• Advancements in CBCT techniques and algorithms require sophisticated simulation capabilities.

Purpose of the Study:

• To develop GECCO, a hybrid GPU-Monte Carlo (MC) software toolkit designed to enhance both speed and accuracy in CBCT simulation.
• Introduce two applications: GECCO for rapid CT to CBCT conversion and RTIS for real-time CBCT image simulation.

Main Methods:

• The GECCO application combines analytical primary projection calculation with GPU-MC for scatter simulation, optimizing the Fastcat toolkit.
• Scatter simulation is enhanced using a modified GGEMs GPU-MC, with denoising and interpolation to reduce computational load.
• The RTIS application employs a purely analytical approach for maximum speed, simulating single CT slices for dynamic, real-time updates.

Main Results:

• GECCO simulated 915 full-resolution CNS patient CBCT projections in 63 minutes on an NVIDIA 4090 GPU.
• GECCO demonstrated an RMSE of 5.7 HU for the Catphan sensitometry module compared to TrueBeam OBI.
• RTIS achieved initialization and update times of 917 ms and 36 ms, respectively, orders of magnitude faster than existing methods, with a 14-16 HU RMSE for CNS patient profiles.

Conclusions:

• The GECCO software toolkit, comprising the GECCO and RTIS applications, offers significant advancements in CBCT simulation.
• GECCO application facilitates the creation of high-quality datasets for machine learning in medical imaging.
• RTIS application enables real-time selection of imaging techniques, benefiting radiation therapists.