Game engine-driven high-precision collision simulation system for radiation therapy: development and clinical evaluation.

Pre-treatment verification is essential to prevent physical collisions between the gantry and the patient during radiation therapy. Conventionally, physical "dry runs" have been performed to ensure safety; however, these procedures can create bottlenecks in clinical workflows. This study aimed to develop virtual collision simulation software using a freely available development environment and to evaluate its performance through phantom studies and verification of clinical treatment plans. High-fidelity three-dimensional (3D) models of the linear accelerator and various immobilization devices were constructed. The software was developed in the Unity environment and incorporated a collision detection system and a function to predict treatment couch coordinates. Collision detection accuracy was assessed using phantom experiments by comparing simulated results with those obtained from the actual linear accelerator. The accuracy of couch coordinate predictions and collision detection performance were retrospectively evaluated using clinical data from 71 treatment plans. All collision prediction errors in the phantom experiments fell conservatively within a 3.00° margin prior to physical contact, and couch coordinate predictions were within 1 cm of the actual values. In the evaluation of clinical plans, the software achieved a true-positive rate of 100% and a negative predictive value of 100%. The developed software achieved high-precision collision detection despite being created in a free platform, thereby ensuring both high performance and broad reproducibility. By serving as an effective pre-screening tool and reducing the need for physical dry runs, this cost-effective and accessible solution streamlines workflow and enhances patient safety.