摘要
针对传统腹腔镜肾部分切除手术术前诊断主观性强、术前术中信息难衔接及术中病灶定位困难导致的手术效率低且存在一定难度和风险的问题,为辅助医生快速准确术前诊断,提出基于级联残差混合卷积网络(RCM UNet)的图像分割算法.首先,通过对CT血管造影(CTA)图像进行分割,分割出左右肾脏的肿瘤、血管和正常组织;然后,利用Marching Cubes算法对肾脏进行重建,更直观地展现患者的病灶信息.在此基础上,利用光学定位系统跟踪腹腔镜相机和术中图像,借助VTK库实现重建模型和术中图像的增强现实叠加,引导医生对肿瘤实施切除,以提高手术精度和效率.实验结果表明:上述方案可实现正常的配准跟踪,在腹腔镜模拟器上可达10帧/s的跟踪速率,且当在开源腹腔镜图像数据集上采集8对以上点时可实现小于3 mm的配准误差,初步验证了本方案的有效性.
Aiming at the problem of the subjective nature of preoperative diagnosis,difficulty in connecting preoperative and intraoperative information,and difficulty in locating lesions during traditional laparoscopic partial nephrectomy resulting in low surgical efficiency and certain difficulties and risks,an image segmentation algorithm based on cascaded residual convolution-mixing network(RCM U-Net)was proposed to assist doctors in rapid and accurate preoperative diagnosis.First,by segmenting CT angiography(CTA)images,tumors,blood vessels,and normal tissues of left and right kidneys could be segmented,and then the Marching Cubes algorithm was used to reconstruct the kidney,presenting patient lesion information more intuitively.On this basis,an optical positioning system was used to track laparoscopic camera and intraoperative images,and the VTK library was utilized to achieve augmented reality overlay of reconstruction models and intraoperative images,which guided doctors to perform tumor resection with improvement of surgical accuracy and efficiency.Experimental results show that this scheme could achieve normal registration and tracking,with 10 frames/s tracking rate on the laparoscopic simulator.When collecting more than 8 pairs of points on the open-source laparoscopic image dataset,a registration error of less than 3 mm could be achieved,which preliminarily verifies the effectiveness of the scheme.
作者
刘冬
张耀仁
任宇
潘德华
赫相
于广海
LIU Dong;ZHANG Yaoren;REN Yu;PAN Dehua;HE Xiang;YU Guanghai(School of Mechanical Engineering,Dalian University of Technology,Dalian 116024,Liaoning China;Ningbo Institute of Dalian University of Technology,Ningbo 315032,Zhejiang China;Dalian Municipal Central Hospital,Dalian 116033,Liaoning China)
出处
《华中科技大学学报(自然科学版)》
北大核心
2025年第11期61-68,共8页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
宁波市“科创甬江2035”关键技术突破计划资助项目(2024Z201)
大连市生命健康领域指导计划资助项目(2022YG013)
辽宁省中央引导地方科技发展专项资助项目(2021JH6/10500144)。
关键词
增强现实
腹腔镜手术
图像分割
图像配准
特征追踪
augmented reality
laparoscopic surgery
image segmentation
image registration
feature tracking
