Development of an Integrated Radiotherapy Simulation Platform with AI-Driven Segmentation and Ray-Casting-Based...

Area of Science:

• Medical Physics
• Artificial Intelligence in Healthcare
• Radiotherapy Technology

Background:

• Accurate radiotherapy simulation is crucial for precise tumor targeting and patient safety.
• Current radiotherapy planning can be time-consuming and costly, necessitating efficiency improvements.

Purpose of the Study:

• To develop an integrated radiotherapy simulation system for automated segmentation, dose estimation, and collision detection.
• To enhance efficiency and reduce costs in radiotherapy treatment planning through a virtual environment.

Main Methods:

• Utilized the Point Transformer model on CT-derived organ point clouds for automated segmentation, with Farthest Point Sampling (FPS) for downsampling.
• Implemented reconstruction and refinement algorithms (k-d tree, outlier removal, marching cubes, surface smoothing) for anatomical fidelity.
• Employed ray casting for beam penetration simulation and dose estimation, integrated with a collision detection module in a Unity 3D environment.

Main Results:

• The Point Transformer model achieved high segmentation accuracy (Dice scores: 93.86% single-organ, 91.86% multi-organ), outperforming PointNet++.
• Ray casting dose estimation showed minimal discrepancies (3.5% brain, 5.9% liver, 13.8% lung) compared to Pinnacle TPS.
• The system successfully integrated segmentation, dose estimation, and collision detection in a virtual simulation environment.

Conclusions:

• The developed integrated radiotherapy simulation system offers a low-cost, adaptable solution for improving radiotherapy planning.
• The AI-driven approach enhances segmentation accuracy and dose estimation, suitable for research, education, and clinical workflow optimization.