Innovative integration of augmented reality and optical surface imaging: A coarse‐to‐precise system for radiotherapy positioning
Gongsen Zhang,Zejun Jiang,Jian Zhu,Tianyuan Dai,Xiaolong He,Xinchao Liu,Yankui Chang,Linlin Wang
DOI: https://doi.org/10.1002/mp.16417
IF: 4.506
2023-04-17
Medical Physics
Abstract:Background Traditional methods of radiotherapy positioning have shortcomings such as fragile skin‐markers, additional doses, and lack of information integration. Emerging technologies may provide alternatives for the relevant clinical practice. Purpose To propose a noninvasive radiotherapy positioning system integrating augmented reality (AR) and optical surface, and to evaluate its feasibility in clinical workflow. Methods AR and structured light‐based surface were integrated to implement the coarse‐to‐precise positioning through two coherent steps, the AR‐based coarse guidance and the optical surface‐based precise verification. To implement quality assurance, recognition of face and pattern was used for patient authentication, case association, and accessory validation in AR scenes. The holographic images reconstructed from simulation computed tomography (CT) images, guided the initial posture correction by virtual‐real alignment. The point clouds of body surface were fused, with the calibration and pose estimation of structured light cameras, and segmented according to the preset regions of interest (ROIs). The global‐to‐local registration for cross‐source point clouds was achieved to calculate couch shifts in six degrees‐of‐freedom (DoF), which were ultimately transmitted to AR scenes. The evaluation based on phantom and human‐body (4 volunteers) included, (i) quality assurance workflow, (ii) errors of both steps and correlation analysis, (iii) receiver operating characteristic (ROC), (iv) distance characteristics of accuracy, and (v) clinical positioning efficiency. Results The maximum errors in phantom evaluation were 3.4 ± 2.5 mm in Vrt and 1.4 ± 1.0° in Pitch for the coarse guidance step, while 1.6 ± 0.9 mm in Vrt and 0.6 ± 0.4° in Pitch for the precise verification step. The Pearson correlation coefficients between precise verification and cone beam CT (CBCT) results were distributed in the interval [0.81, 0.85]. In ROC analysis, the areas under the curve (AUC) were 0.87 and 0.89 for translation and rotation, respectively. In human body‐based evaluation, the errors of thorax and abdomen (T&A) were significantly greater than those of head and neck (H&N) in Vrt (2.6 ± 1.1 vs. 1.7 ± 0.8, p
radiology, nuclear medicine & medical imaging