An Investigation of Two Cost-Effective 3D Optical Sensing Technologies for Real-Time Collision and Monitoring for Proton Therapy.
Abstract
Purpose
This study investigates the feasibility of using two affordable optical surface–imaging technologies to generate detailed 3D maps of the treatment room and patient surface for detecting and mitigating potential collisions between the proton nozzle and the patient during treatment delivery.
Methods
Two optical imaging systems were evaluated: the Livox Mid‑70 LiDAR, which uses a Class‑I 905 nm laser with an operating range of 0.2-260m and wide circular field‑of‑view (FoV), and the Helios2+ Time‑of‑Flight (ToF) camera, which employs Class‑I 850 nm laser diodes with an operating range of 0.3–8.0 m. Each system was used to acquire 3D point-clouds of a gantry‑mounted proton therapy vault and a volunteer. Point cloud data were processed using the CloudCompare software package for segmentation and object identification.
Results
Both systems successfully generated 3D scans of the treatment vault and subject. The LiDAR system captured a wide 70.4° FoV and provided highly detailed room‑scale geometry, though the scans exhibited increased noise and flying voxels due to its limited precision (20 mm) at long operating ranges. In contrast, the ToF camera produced cleaner and less noisy scans of subject contours due to higher surface accuracy (±10 mm) but was constrained by its narrower 69° × 51° FoV and lower 640 × 480 pixel resolution. Both systems showed difficulty imaging transparent surfaces (e.g., nozzle window) and very dark materials (e.g., couch top). Multiple scans demonstrated consistent reconstruction of gantry and patient geometry relevant for collision assessment.
Conclusion
Low‑cost LiDAR and ToF imaging technologies can produce sufficiently detailed 3D maps for room‑scale and patient‑surface modeling in proton therapy. The LiDAR system is suitable for comprehensive room mapping, whereas the ToF camera provides more accurate patient‑surface capture. Their robustness to ambient light and ability to continuously scan during treatment support their potential use in an affordable, real‑time collision detection system.