| 基于光学体表监控和X射线透视影像的膈肌运动自动跟踪 |
| |
| 引用本文： | 谭翔,戴振晖,何强,张白霖,朱琳,蔡春雅,杨耕,简婉薇,王学涛.基于光学体表监控和X射线透视影像的膈肌运动自动跟踪[J].中国医学物理学杂志,2022,0(12):1453-1459. |
| |
| 作者姓名： | 谭翔 戴振晖 何强 张白霖 朱琳 蔡春雅 杨耕 简婉薇 王学涛 |
| |
| 作者单位： | 广州中医药大学第二附属医院放射治疗区， 广东 广州 510006 |
| |
| 摘 要： | 目的:基于直线加速器的光学体表监控系统和X射线透视影像利用人工智能构建膈肌顶点运动的自动跟踪模型。方法:同步采集7例肝肿瘤患者胸腹部的光学体表运动信息和千伏级X射线透视影像，选取其中3例患者数据利用主成分分析与偏最小二乘回归结合的方法计算不同体表感兴趣区域与膈肌运动的相关系数，选择相关系数最大的体表感兴趣区域作为光学体表监控区。首先，使用全卷积网络模型自动识别透视图像中膈肌顶点的位置；再利用随机森林方法建立体表与膈肌顶点运动的关联模型，基于体表运动信息实时预测膈肌顶点运动轨迹；最后，把自动跟踪的膈肌顶点位置与放疗医生手动勾画位置进行对比，以评估模型精度。结果:3例患者的体表感兴趣区域与膈肌运动的平均相关系数在前后（AP）方向最高达到（0.73±0.01） mm，上下（SI）方向最高达到（0.88±0.01） mm。自动跟踪模型预测结果与手动勾画位置的平均绝对误差和均方根误差SI方向分别为（3.09±0.79） mm和（3.89±0.89） mm，AP方向分别为（1.42±0.43） mm和（1.78±0.46） mm。结论:体表呼吸运动与体内膈肌运动是相关的，在放疗过程中基于光学体表运动信息可以实时跟踪体内膈肌顶点运动，该技术可用于胸腹部肿瘤放疗期间膈肌附近肿瘤的实时及无创运动管理。
|
| 关 键 词： | 膈肌跟踪 人工智能 光学体表监控系统 透视影像 |
Automated diaphragm motion tracking using optical surface monitoring system and X-ray fluoroscopic image |
| |
| TAN Xiang,DAI Zhenhui,HE Qiang,ZHANG Bailin,ZHU Lin,CAI Chunya,YANG Geng,JIAN Wanwei,WANG Xuetao.Automated diaphragm motion tracking using optical surface monitoring system and X-ray fluoroscopic image[J].Chinese Journal of Medical Physics,2022,0(12):1453-1459. |
| |
| Authors: | TAN Xiang DAI Zhenhui HE Qiang ZHANG Bailin ZHU Lin CAI Chunya YANG Geng JIAN Wanwei WANG Xuetao |
| |
| Affiliation: | Department of Radiation Therapy, the Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou 510006, China |
| |
| Abstract: | Abstract: Objective To establish an automated diaphragm apex motion tracking model with artificial intelligence based on the optical surface monitoring system of linear accelerator and X-ray fluoroscopic image. Methods The optical surface motion information and kV X-ray fluoroscopic images of the thoracic and abdominal regions in 7 patients with liver tumors were acquired synchronously. The principal component analysis combined with partial least squares regression was used to calculate the correlation coefficients between several body surface regions of interest and diaphragm motion in 3 patients from 7 patients, and the body surface region of interest with the largest correlation coefficient was selected as the optical surface monitoring area. After automatically identifying the position of the diaphragm apex in the fluoroscopic images using fully convolutional neural network model, the correlation model between the body surface and the diaphragm apex motion was established with random forest method to predict the trajectory of the diaphragm apex in real time based on the body surface motion information. The accuracy of the established model was assessed by comparing the automatically tracked diaphragm apex position with the position manually drawn by the radiation oncologist. Results The mean correlation coefficient between body surface regions of interest and diaphragm motion in the 3 patients reached a maximum of (0.73±0.01) mm in the anterior-posterior direction, and a maximum of (0.88±0.01) mm in the superior-inferior direction. The mean absolute error and root mean square error between the predicted results of the automated tracking model and the manually delineated position were (3.09±0.79) mm and (3.89±0.89) mm in superior-inferior direction, (1.42±0.43) mm and (1.78±0.46) mm in anterior-posterior direction. Conclusion The body surface respiratory motion is associated with the internal diaphragm motion. The diaphragm apex motion can be tracked in real time using the optical surface motion information during radiotherapy, and the technique can be used for real time and non-invasive motion management of tumor near the diaphragm during radiotherapy of thoracic and abdominal tumors. |
| |
| Keywords: | Keywords: diaphragm tracking artificial intelligence optical surface monitoring system fluoroscopic image |
|
| 点击此处可从《中国医学物理学杂志》浏览原始摘要信息 |
|
点击此处可从《中国医学物理学杂志》下载全文 |
