Optimizing dosimetry in Y-90 microsphere radioembolization: GPU-accelerated Monte Carlo simulation versus...

Area of Science:

• Medical Physics
• Interventional Oncology
• Radiotherapy Dosimetry

Background:

• Yttrium-90 (90Y) microsphere radioembolization is a key treatment for liver cancer.
• Current dosimetric models face challenges in clinical implementation, especially in high-volume settings.
• There is a need for accurate and efficient dosimetry to optimize 90Y radioembolization workflows.

Purpose of the Study:

• To develop a GPU-based fast Monte Carlo simulation for accurate voxel-level dose calculation in 90Y radioembolization.
• To evaluate the performance of the GPU-based approach against existing dosimetric methods.
• To support optimized clinical workflows in Chinese interventional oncology departments.

Main Methods:

• A GPU-accelerated fast Monte Carlo simulation algorithm was developed.
• The algorithm was retrospectively applied to eight patients with liver cancer.
• Dosimetric performance was compared against direct Monte Carlo, MIRD, VSV, and LED models using MARE and RSD metrics.

Main Results:

• The GPU-based Monte Carlo code demonstrated high accuracy and efficiency, with superior voxel-level consistency (median RSD -1.13%) compared to other methods.
• It achieved significantly lower Mean Absolute Relative Error (MARE) at 4.53% versus VSV (6.71%) and LED (49.36%).
• The GPU method showed a 1,296-fold speed-up over traditional Monte Carlo simulations, enabling efficient clinical dosimetry.

Conclusions:

• GPU-based fast Monte Carlo simulation is a highly accurate and efficient tool for voxel-level dosimetry in 90Y radioembolization.
• This method allows precise dose estimation, minimizing risks like radiation pneumonitis and supporting high-throughput workflows.
• A stratified approach, using simplified methods for simple cases and GPU-based MC for complex ones, can optimize accuracy and efficiency.