基于振动信号融合的手术机器人椎板磨削剩余厚度识别  

Recognition of Remaining Thickness of Lamina Milling via a Surgical Robot Based on Vibration Signal Fusion

作　　者：夏光明 王瑞 张丽娜 张建勋 代煜 Xia Guangming;Wang Rui;Zhang Lina;Zhang Jianxun;Dai Yu(College of Artificial Intelligence,Nankai University,Tianjin 300350,China;Department of Orthopaedics Surgery,Tianjin Medical University General Hospital,Tianjin 300052,China)

机构地区：[1]南开大学人工智能学院,天津300350 [2]天津医科大学总医院骨科,天津300052

出　　处：《天津大学学报（自然科学与工程技术版）》2022年第10期1016-1025,共10页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(61773223,U1913207).

摘　　要：椎板切除术是一种用于治疗椎管狭窄症的骨外科手术,通过移除椎板来恢复椎管空间和解除脊髓压迫.椎板磨削是椎板切除术中的核心环节.在机器人磨薄椎板的过程中,磨钻需要在接近脊髓的1~2 mm左右的区域内工作,存在较高手术风险.使用脊柱手术机器人磨削椎板的过程中的关键问题之一是如何在术中估计椎板的剩余厚度来决定是否停止磨削操作.为解决上述问题,本文首先分析了机器人对椎板的逐层磨削过程,通过建立椎板磨削振动模型,给出了根据磨钻切入椎板过程中的振动信号来估计椎板剩余厚度的原理.然后搭建了脊柱手术机器人样机,使用机器人按规划轨迹逐层切入和磨薄猪脊骨的椎板,并使用振动传感器和激光位移传感器分别采集手术磨钻切入椎板过程中磨钻和椎板在切深方向上的振动信号.最后计算两种振动信号中对应磨钻旋转频率的谐波幅值和相对幅值来构造特征向量和训练神经网络,用于识别椎板剩余厚度.实验结果表明,考虑相对幅值的椎板剩余厚度的识别成功率更高.磨削力扰动下,3 mm、2 mm和1 mm的成功率分别为96.7%、96.9%和100%.92组实验中,仅有1组3 mm被识别为2 mm,所有磨钻切入2 mm和1 mm椎板的过程均被准确识别.所提方法有利于提升脊柱手术机器人自动椎板磨削过程的智能化程度和安全性.Laminectomy is a bone surgery that removes the lamina to restore spinal canal space and relieve spinal cord compression in patients with spinal stenosis.The critical operation of laminectomy is lamina milling.The milling tool must work in an area of approximately 1—2 mm close to the spinal cord when thinning the lamina with a robot,posing a higher surgical risk.A critical issue in milling the lamina using a spinal surgical robot is determining how to estimate the remaining thickness of the lamina intraoperatively to decide whether to stop the milling operation.To address the aforementioned issues,the lamina layer-by-layer milling process implemented by the robot is first examined,and it is demonstrated that the remaining thickness of the lamina can be estimated using vibration signals generated when the cutter mills into the lamina by establishing the lamina-milling vibration model.Thereafter,a prototype of the spinal surgical robot was built,which was used to cut and thin the laminas of the porcine spine layer-by-layer according to the planned trajectory.Vibration and laser displacement sensors were used to collect the vibration signals of the cutter and lamina in the cutting depth direction during the process of the cutter milling into the lamina.Finally,the harmonic and the relative amplitudes of these two vibration signals,corresponding to the rotation frequency of the cutter,were calculated to construct the feature vector and train the neural network to identify the remaining thickness of the lamina.The experimental results show that the recognition success rate of the remaining thickness of the lamina considering the relative amplitude is the highest,with success rates for 3,2,and 1 mm being 96.7%,96.9%,and 100%,respectively.In 92 groups of experiments,only one group of 3 mm was identified as 2 mm,and all the processes of the cutter cutting into 2 and 1 mm lamina were correctly identified.The proposed method is beneficial for improving the intelligence and the safety of the automatic lamina milling operat

关 键 词：脊柱手术机器人 椎板切除术 椎板磨削 振动信号融合 神经网络 状态识别 

分 类 号：TP242.3[自动化与计算机技术—检测技术与自动化装置]