Collision detection in radiotherapy: A comprehensive software approach using depth camera and separating axis theorem

Purpose

Collision detection in radiotherapy is essential for ensuring patient and equipment safety, as well as accurate treatment planning in personalized medicine. Despite the use of touch guards on modern linear accelerators (LINACs) and extensive research, shortcomings in efficiency and versatility still remain. This study presents a patient-specific collision detection software that supports a variety of LINACs with complex motion trajectories.

Methods

A depth camera was utilized to obtain point clouds of objects, while the hand-eye calibration technique was applied for model localization. Analysis of the equipment structure and radiation therapy planning files (RT files) enabled the extraction of motion trajectories and relevant parameters for testing. Our software integrates the separating axis theorem (SAT) algorithm with the concept of collision bodies to simplify collision detection complexity. The axis-aligned bounding box (AABB), bounding volume hierarchy (BVH), and collision pair analysis were applied to optimize the simulation efficiency.

Results

The proposed method exhibited robust performance in simulating and detecting collisions during the radiotherapy process. By applying the concepts of bounding volumes (BVs) and BVH, the computational complexity of intricate components, such as the radiation delivery and imaging system (RDIS) and patient positioning system (PPS), was reduced by over 95 % and 99 %, respectively. This significantly enhanced the efficiency of collision detection.

Conclusion

This study enhances safety and operational efficiency in radiotherapy, providing a reliable tool for clinical use. Additionally, it offers interactive education and training opportunities for novice healthcare professionals, bridging gaps in both practice and learning.