融合形态特征的基于GRU的介入机器人导丝轨迹预测建模

张任飞; 董林杰; 王兴松; 田梦倩; 苏浩波

doi:10.16383/j.aas.c250506

融合形态特征的基于GRU的介入机器人导丝轨迹预测建模

doi: 10.16383/j.aas.c250506 cstr: 32138.14.j.aas.c250506

1.
东南大学机械工程学院南京 211189
2.
南京市第一医院介入血管科南京 210006

基金项目: 国家自然科学基金(52175005)资助

详细信息

作者简介:
张任飞：东南大学机械工程学院博士研究生. 主要研究方向为介入手术机器人的导航与控制. E-mail: 230218562@seu.edu.cn

董林杰：东南大学机械工程学院博士研究生. 主要研究方向为基于深度学习的图像处理与人工智能. E-mail: 230228047@seu.seu.cn

王兴松：博士, 东南大学机械工程学院教授. 主要研究方向为机器人动力学及其控制, 先进医疗器械. 本文通信作者. E-mail: xswang@seu.edu.cn

田梦倩：博士, 东南大学机械工程学院副教授. 主要研究方向为智能机器人, 控制技术, 机电控制及自动化. E-mail: Tianmq@seu.edu.cn

苏浩波：博士, 南京市第一医院介入血管科副主任医师. 主要研究方向为脑血管疾病及良恶性肿瘤的介入诊疗. E-mail: doctorsuhaobo@163.com

计量
- 文章访问数: 263
- HTML全文浏览量: 143
- PDF下载量: 69
- 被引次数: 0
出版历程
- 收稿日期: 2025-09-29
- 录用日期: 2025-12-19
- 网络出版日期: 2026-02-12
- 刊出日期: 2026-03-20

GRU-based Modeling for Predicting Guidewire Trajectories in Interventional Robotics With Morphological Feature Fusion

1.
School of Mechanical Engineering, Southeast University, Nanjing 211189
2.
Department of Interventional and Vascular Radiology, Nanjing First Hospital, Nanjing 210006

Funds: Supported by National Natural Science Foundation of China (52175005)

More Information

Author Bio:
ZHANG Ren-Fei　Ph.D. candidate at the School of Mechanical Engineering, Southeast University. His main research interest is navigation and control of interventional surgical robots

DONG Lin-Jie　Ph.D. candidate at the School of Mechanical Engineering, Southeast University. His research interests include deep-learning-based image processing and artificial intelligence

WANG Xing-Song　Ph.D., professor at the School of Mechanical Engineering, Southeast University. His research interests include robot dynamics and control, and advanced medical devices. Corresponding author of this paper

TIAN Meng-Qian　Ph.D., associate professor at the School of Mechanical Engineering, Southeast University. Her research interests include intelligent robots, control technology, and mechatronic control and automation

SU Hao-Bo　Ph.D., associate chief physician in the Department of Interventional and Vascular Radiology, Nanjing First Hospital. His research interests include interventional diagnosis and treatment of cerebrovascular diseases and benign and malignant tumors

摘要

摘要: 针对介入导航场景中的导丝轨迹重建问题, 提出一种保持因果性的序列估计方法. 不同于通用循环基线模型, 所提方法将序列级常量特征 (包括导丝刚度、进入角度和有效摩擦描述符) 按时间步广播, 与动态几何量(中心线坐标、直径等)拼接, 经两层特征编码后由单向门控循环单元解码器逐时输出二维坐标. 为处理变长序列, 本文采用时间步长分类的训练策略, 并结合掩码损失函数, 以抑制填充引入的无效梯度, 在不改变网络结构的前提下提升训练与推理效率. 基于覆盖多类导丝与多进入角度的仿体实验平台, 所提方法在保持因果性的同时, 实现0.40 ~ 0.54 mm的位置误差范围(平均误差为0.46 mm); 相较于未采用时间步分类策略的基线模型, 收敛epoch降低42%, 训练时间降低52%, 单次推理时延降低51%. 结果表明, 该方法可为导丝轨迹估计与术中导航提供可部署的算法基础.
- 介入机器人 /
- 导丝轨迹预测 /
- 门控循环单元 /
- 形态特征融合 /
- 时序建模
Abstract: We present a causality-preserving sequential estimator for guidewire trajectory reconstruction during interventional navigation. Unlike generic recurrent baselines, the proposed model time-broadcasts sequence-level constants (including guidewire stiffness, insertion angle, and an effective friction descriptor) and concatenates them with dynamic geometric tokens (centerline coordinates and local diameter) before a two-stage feature encoder and a unidirectional gated recurrent unit decoder that emits 2D positions stepwise. To cope with variable sequence lengths, we adopt a time-step length classification training strategy with mask-based loss function, which limits padding-induced invalid gradients and improves training and inference efficiency without altering the network architecture. On a phantom platform covering multiple guidewire types and insertion angles, the method achieves a 0.40 ~ 0.54 mm position-error range (mean 0.46 mm) while preserving strict causality; relative to a baseline without the time-step classification strategy, it reduces epochs-to-convergence by 42%, training time by 52%, and per-inference latency by 51%. These results indicate a deployable, real-time basis for guidewire trajectory estimation and intraoperative navigation.
- interventional robotics /
- guidewire trajectory prediction /
- gated recurrent unit /
- morphological feature fusion /
- sequential modeling

HTML全文

图 1 血管内介入与细长机器人辅助引导系统

Fig. 1 Endovascular intervention and slender robot assisted guidance system

下载: 全尺寸图片幻灯片

图 2 导丝形状预测流程示意图

Fig. 2 Schematic illustration of the guidewire shape prediction pipeline

下载: 全尺寸图片幻灯片

图 3 基于GRU的导丝整体轨迹预测模型

Fig. 3 GRU-based model for whole guidewire trajectory prediction

下载: 全尺寸图片幻灯片

图 4 血管结构的形态特征测量结果示意图

Fig. 4 Illustration of morphological feature measurements results for vascular structures

下载: 全尺寸图片幻灯片

图 5 用户实验平台的组成结构示意图

Fig. 5 Schematic diagram of the user experiment platform and its components

下载: 全尺寸图片幻灯片

图 6 多目标任务中导丝轨迹真实与预测对比图

Fig. 6 Comparison of actual and predicted guidewire trajectories in multi-target tasks

下载: 全尺寸图片幻灯片

图 7 导丝插入任务中实拍图像上模型预测轨迹可视化

Fig. 7 Visualization of model-predicted trajectories overlaid on real captured images in guidewire insertion tasks

下载: 全尺寸图片幻灯片

图 8 不同导丝与插入角度组合下的MAE和CV

Fig. 8 MAE and CV across different guidewire and insertion angle configurations

下载: 全尺寸图片幻灯片

表 1 时间步分类策略对模型实时性的影响

Table 1 Impact of the time-step classification strategy on model real-time performance

是否使用时间步
分类策略 epoch数量训练用时
(h) 单次推理时延
(ms)

是 377 0.221 11.571

否 653 0.457 23.783

下载: 导出CSV

参考文献(32)

[1]	Robertshaw H, Karstensen L, Jackson B, Granados A, Booth T C. Autonomous navigation of catheters and guidewires in mechanical thrombectomy using inverse reinforcement learning. International Journal of Computer Assisted Radiology and Surgery, 2024, 19(8): 1569−1578 doi: 10.1007/s11548-024-03208-w
[2]	奉振球, 侯增广, 边桂彬, 谢晓亮, 周小虎. 微创血管介入手术机器人的主从交互控制方法与实现. 自动化学报, 2016, 42(5): 696−705 doi: 10.16383/j.aas.2016.c150577 Feng Zhen-Qiu, Hou Zeng-Guang, Bian Gui-Bin, Xie Xiao-Liang, Zhou Xiao-Hu. Master-slave interactive control and implementation for minimally invasive vascular interventional robots. Acta Automatica Sinica, 2016, 42(5): 696−705 doi: 10.16383/j.aas.2016.c150577
[3]	Li L, Li X J, Yang S L, Ding S, Jolfaei A, Zheng X. Unsupervised-learning-based continuous depth and motion estimation with monocular endoscopy for virtual reality minimally invasive surgery. IEEE Transactions on Industrial Informatics, 2021, 17(6): 3920−3928 doi: 10.1109/TII.2020.3011067
[4]	Chen T, Zhao X W, Zhang Y H, Zheng G, Hou L C, Ling Q, et al. Ultrasound-guided robotic autonomous operation based on real-time deformation tracking and prediction. IEEE Transactions on Industrial Informatics, 2025, 21(2): 1369−1378 doi: 10.1109/TII.2024.3477549
[5]	张天, 刘检华, 唐承统, 刘少丽. 基于中心线匹配的导管三维重建技术. 自动化学报, 2015, 41(4): 735−748 doi: 10.16383/j.aas.2015.c130287 Zhang Tian, Liu Jian-Hua, Tang Cheng-Tong, Liu Shao-Li. 3D reconstruction technique of pipe based on centerline matching. Acta Automatica Sinica, 2015, 41(4): 735−748 doi: 10.16383/j.aas.2015.c130287
[6]	金朝勇, 耿国华, 李姬俊男, 周明全, 朱新懿. 一种新的虚拟血管镜自动导航路径生成方法. 自动化学报, 2015, 41(8): 1412−1418 doi: 10.16383/j.aas.2015.c150014 Jin Chao-Yong, Geng Guo-Hua, Li Ji-Jun-Nan, Zhou Ming-Quan, Zhu Xin-Yi. A new automatic navigation path generation approach to virtual angioscopy. Acta Automatica Sinica, 2015, 41(8): 1412−1418 doi: 10.16383/j.aas.2015.c150014
[7]	张名洋, 边桂彬, 李桢, 叶强. 支气管介入的多段柔性连续体机器人实时形状重建. 科学技术与工程, 2025, 25(1): 245−251 Zhang Ming-Yang, Bian Gui-Bin, Li Zhen, Ye Qiang. Real-time shape reconstruction of flexible multi-segment continuum robot based for bronchial intervention. Science Technology and Engineering, 2025, 25(1): 245−251
[8]	Pore A, Li Z, Dall'Alba D, Hernansanz A, de Momi E, Menciassi A, et al. Autonomous navigation for robot-assisted intraluminal and endovascular procedures: A systematic review. IEEE Transactions on Robotics, 2023, 39(4): 2529−2548 doi: 10.1109/TRO.2023.3269384
[9]	Li P, Feng J, Zhang X, Fang D L, Zhang J X, Liang C M. Modeling and experimental study of the intervention forces between the guidewire and blood vessels. Medical Engineering & Physics, 2024, 127: Article No. 104166 doi: 10.1016/j.medengphy.2024.104166
[10]	Ravigopal S R, Brumfiel T A, Sarma A, Desai J P. Fluoroscopic image-based 3-D environment reconstruction and automated path planning for a robotically steerable guidewire. IEEE Robotics and Automation Letters, 2022, 7(4): 11918−11925 doi: 10.1109/LRA.2022.3207568
[11]	Zhang R F, Dong L J, Wang X S, Tian M Q, Su H B. A sensor-less guider contact force estimation approach for endovascular slender robot-assisted guidance system. IEEE Transactions on Instrumentation and Measurement, 2025, 74: 1−9 doi: 10.1109/tim.2025.3545167
[12]	Zhang R F, Dong L J, Wang X S, Tian M Q, Su H B. Morphological feature measurement of specific locations using image processing. In: Proceedings of the 8th International Conference on Machine Vision and Applications (ICMVA 2025). Melbourne, Australia: SPIE, 2025, 13734: 9−17
[13]	Shen D H, Zhang Q, Han Y L, Tu C L, Wang X S. Design and development of a continuum robot with switching-stiffness. Soft Robotics, 2023, 10(5): 1015−1027 doi: 10.1089/soro.2022.0179
[14]	Sharei H, Alderliesten T, van den Dobbelsteen J J, Dankelman J. Navigation of guidewires and catheters in the body during intervention procedures: A review of computer-based models. Journal of Medical Imaging, 2018, 5(1): Article No. 010902 doi: 10.1117/1.jmi.5.1.010902
[15]	Beasley R A, Howe R D. Increasing accuracy in image-guided robotic surgery through tip tracking and model-based flexion correction. IEEE Transactions on Robotics, 2009, 25(2): 292−302 doi: 10.1109/TRO.2009.2014498
[16]	蒋芸, 谭宁. 基于条件深度卷积生成对抗网络的视网膜血管分割. 自动化学报, 2021, 47(1): 136−147 Jiang Yun, Tan Ning. Retinal vessel segmentation based on conditional deep convolutional generative adver-sarial networks. Acta Automatica Sinica, 2021, 47(1): 136−147
[17]	Tai Y H, Qian K, Huang X Q, Zhang J, Jan M A, Yu Z T. Intelligent intraoperative haptic-AR navigation for COVID-19 lung biopsy using deep hybrid model. IEEE Transactions on Industrial Informatics, 2021, 17(9): 6519−6527 doi: 10.1109/TII.2021.3052788
[18]	Zhao Y, Wang Y X, Zhang J H, Liu X K, Li Y X, Guo S X, et al. Surgical GAN: Towards real-time path planning for passive flexible tools in endovascular surgeries. Neurocomputing, 2022, 500: 567−580 doi: 10.1016/j.neucom.2022.05.044
[19]	Qi F, Ju F, Bai D M, Wang Y Y, Chen B. Kinematic analysis and navigation method of a cable-driven continuum robot used for minimally invasive surgery. The International Journal of Medical Robotics and Computer Assisted Surgery, 2019, 15(4): Article No. e2007 doi: 10.1002/rcs.2007
[20]	Li Z, Dankelman J, de Momi E. Path planning for endovascular catheterization under curvature constraints via two-phase searching approach. International Journal of Computer Assisted Radiology and Surgery, 2021, 16(4): 619−627 doi: 10.1007/s11548-021-02328-x
[21]	Gao M K, Chen Y M, Liu Q, Huang C, Li Z Y, Zhang D H. Three-dimensional path planning and guidance of leg vascular based on improved ant colony algorithm in augmented reality. Journal of Medical Systems, 2015, 39(11): Article No. 133 doi: 10.1007/s10916-015-0315-2
[22]	Scarponi V, Duprez M, Nageotte F, Cotin S. A zero-shot reinforcement learning strategy for autonomous guidewire navigation. International Journal of Computer Assisted Radiology and Surgery, 2024, 19(6): 1185−1192 doi: 10.1007/s11548-024-03092-4
[23]	Kweon J, Kim K, Lee C, Kwon H, Park J, Song K, et al. Deep reinforcement learning for guidewire navigation in coronary artery phantom. IEEE Access, 2021, 9: 166409−166422 doi: 10.1109/ACCESS.2021.3135277
[24]	Vy P, Auffret V, Castro M, Badel P, Rochette M, Haigron P, et al. Patient-specific simulation of guidewire deformation during transcatheter aortic valve implantation. International Journal for Numerical Methods in Biomedical Engineering, 2018, 34(6): Article No. e2974 doi: 10.1002/cnm.2974
[25]	Cheng X R, Song Q K, Xie X L, Cheng L, Wang L, Bian G B, et al. A fast and stable guidewire model for minimally invasive vascular surgery based on Lagrange multipliers. In: Proceedings of the 7th International Conference on Information Science and Technology (ICIST). Da Nang, Vietnam: IEEE, 2017. 109−114
[26]	Schafer S, Singh V, Noël P B, Walczak A M, Xu J, Hoffmann K R. Real-time endovascular guidewire position simulation using shortest path algorithms. International Journal of Computer Assisted Radiology and Surgery, 2009, 4(6): 597−608 doi: 10.1007/s11548-009-0385-z
[27]	Qiu J P, Zhang L B, Yang G Y, Chen Y, Zhou S J. An improved real-time endovascular guidewire position simulation using activity on edge network. IEEE Access, 2019, 7: 126618−126624 doi: 10.1109/ACCESS.2019.2935327
[28]	Li N, Wang Y W, Zhao H, Ding H. Sensor-free strategy for estimating guidewire/catheter shape and contact force in endovascular interventions. IEEE Robotics and Automation Letters, 2025, 10(1): 264−271 doi: 10.1109/LRA.2024.3504236
[29]	Chen K, Qin W J, Xie Y Q, Zhou S J. Towards real time guide wire shape extraction in fluoroscopic sequences: A two phase deep learning scheme to extract sparse curvilinear structures. Computerized Medical Imaging and Graphics, 2021, 94: Article No. 101989 doi: 10.1016/j.compmedimag.2021.101989
[30]	Ha X T, Wu D, Ourak M, Borghesan G, Dankelman J, Menciassi A, et al. Shape sensing of flexible robots based on deep learning. IEEE Transactions on Robotics, 2023, 39(2): 1580−1593 doi: 10.1109/TRO.2022.3221368
[31]	黎定佳, 郭钰琪, 杨永明, 李锐, 刘连庆, 刘浩. 腔道介入交互约束下的柔性机器人具身形态感知方法. 机器人, 2025, 47(4): 497−507 doi: 10.13973/j.cnki.robot.250202 Li Ding-Jia, Guo Yu-Qi, Yang Yong-Ming, Li Rui, Liu Lian-Qing, Liu Hao. Embodied morphological perception for flexible robots with interactive constraints in luminal intervention. Robot, 2025, 47(4): 497−507 doi: 10.13973/j.cnki.robot.250202
[32]	Kuntz A, Sethi A, Webster R J, Alterovitz R. Learning the complete shape of concentric tube robots. IEEE Transactions on Medical Robotics and Bionics, 2020, 2(2): 140−147 doi: 10.1109/TMRB.2020.2974523