Purpose: Currently commercially available surface guided radiation therapy (SGRT) device is usually fixed on ceil or mounted on gantry. This has limited the lay-out and usability. On the other hand, a portable SGRT system allows SGRT to be performed from any camera angle and position. The purpose of this study is to develop a potable body surface monitoring system and to evaluate the system for clinical implementation.
Methods: We developed a body surface monitoring system that was consisted of Kinect (Azure Kinect DK, Microsoft), which has RGB and depth camera, and software programmed by Python. Coordinate systems of our system and treatment room were aligned by the calibration using RGB images of chessboard. Depth map of the phantom captured by Kinect were converted into 3D point cloud. Finally, the phantom movement was calculated by the rigid registration using point-to-point iterative closest point (p2p-ICP) and point-to-plane ICP (p2pl-ICP) algorithms. The phantom movement were derived from the transformation matrix by two ICP algorithms. Positional detectability was evaluated using general head and neck phantom. The phantom was moved in the range of 10 to 20 mm in the x, y, and z directions, respectively, using the treatment couch (Imaging Couch, Brainlab). The detection accuracy was compered between two ICP algorithms.
Results: As results of detectability evaluation, translational error of p2p-ICP was not satisfied the tolerance of 2 mm defined by AAPM TG-147 in 2 conditions. On the other hand, translational error of p2pl-ICP was satisfied the tolerance in all conditions. The largest error of p2p-ICP and p2pl-ICP were 2.8 mm in z direction and 1.4 mm in x direction, respectively.
Conclusion: We developed the portable body surface monitoring system and evaluated the basic performance. These results show our system with p2pl-ICP algorithm has the potential for clinically implementation.