毕业生知识库

开颅是治疗颅脑创伤、胶质瘤等神经外科疾病的第一步。传统开颅手术存在医生劳动强度大，硬脑膜易损伤，培养合格的神经外科医生周期长，基层医院因条件限制无法及时实施开颅手术导致患者无法获得及时治疗等问题。因此面向神经外科临床应用的机器人控制技术研究具有重要的社会效益和临床应用价值。利用机器人技术辅助开颅能提高手术操作的精准性、稳定性，缩短手术时长，降低医生的劳动消耗，提高手术效率。本文在国家自然科学基金委员会国家重大科研仪器研制项目“面向神经外科肿瘤切除的高灵敏性智能显微精准导航操作仪"（62027813）的支持下，以提高神经外科机器人开颅操作的智能化程度为出发点，研究机器人辅助开颅手术中颅骨钻削和铣削操作中的精准控制问题。论文的主要内容如下：

1. 针对如何让机器人自主执行颅骨钻孔任务的问题，设计了机器人钻孔的进给力、进给速度、姿态混合约束控制方法，并搭建了开颅手术机器人机器人实验平台对控制方法进行验证。本文搭建的开颅手术机器人实验平台搭载了力感知模块模拟医生对力信号变化的感知。在机器人自主颅骨钻孔控制方法上，将机器人钻孔的运动约束成一条直线并通过一个起停式（bang-bang）控制器在进给力过大时限制机械臂钻进的速度来控制进给力。通过离体动物的颅骨钻孔实验的进给力信号，可以将钻孔状态分类为：进给力上升阶段、力饱和阶段、突破阶段和机械保护阶段等四个阶段。

2. 针对机器人在开颅钻削过程中如何自主识别颅骨钻穿问题，本文设计了基于监测进给力信号变化的颅骨-脑膜边界检测算法。该方法设置了两个连续时间序列窗口，第一个连续时间序列窗口T1用于跟踪进给力的变化，同时均值起到减少噪声影响的作用。第一个连续时间序列窗口T2用于缓冲，减少突破阶段的数据进入T1影响检测。针对如何合理选取颅骨-脑膜边界检测算法的参数的问题，本文建立了开颅钻突破阶段进给力-突入量关系的线性近似模型，将开颅钻的额外突入量作为优化目标，将实验中力饱和阶段噪声的统计数据作为约束将颅骨-脑膜边界检测算法的参数选择转化成一个最优化问题，通过求解该优化问题设置了颅骨-脑膜边界检测算法的参数。本研究共利用机器人开颅手术平台进行了72次机器人自主钻孔、自主检测颅骨-脑膜边界的钻孔实验，实验分别在七个山羊头、一个比格犬、一个柴犬的离体头颅上进行。其中有四例实验颅骨没有被钻穿而开颅钻的机械保护装置提前触发，实验被迫终止；一例实验颅骨已钻穿，但是手术机器人未能识别到信号；手术机器人实验平台能自行识别到颅骨被钻穿的实验有67次，占比93.06%，实验均未观测到颅内软组织损伤。

3. 针对机器人自主铣削颅骨的问题设计了基于PID模糊控制的机器人铣削颅骨的力位混合控制方法。本文将上拉力的产生建模为铣刀保护鞘“J”形结构的弹性形变，并结合机械臂与未知环境的阻抗模型提出机械臂装载铣刀上拉力的控制可以利用时变、与位置非线性相关的PD控制器进行控制。为了处理模型不确定与环境未知的力位混合控制问题，本文利用模糊PID对上拉力进行控制，并分别在塑料颅骨模型和离体柴犬颅骨上进行了铣削实验。上拉力控制在塑料颅骨模型实验上的控制均方根误差为1.24 N。动物颅骨内表面并不光滑，存在许多突起与沟壑；离体柴犬头颅保留了相对完整的生物学解剖结构，在对离体的柴犬颅骨铣削过程中，硬脑膜一直与铣刀保护鞘保持弹性接触，实验表明所设计的控制方法虽然在铣削动物颅骨过程中对铣刀上拉力的控制存在波动，但是在铣削过程中仍保持了上拉力，铣削实验未损伤离体动物颅内软组织。

Craniotomy is the first step in the treatment of neurosurgical diseases such as craniocerebral trauma and glioma. Traditional craniotomy has problems such as high labor intensity for sugeons, easy damage to the dura mater, long period of training qualified neurosurgeons, and lack of timely craniotomy in grassroots hospitals due to limited conditions, resulting in patients not being able to receive timely treatment. Therefore, the study on robot control technology for neurosurgery clinical application has important social benefits and clinical application value. The use of robotic technology to assist craniotomy can improve the accuracy and stability of surgical operations, shorten the duration of surgery, reduce the labor consumption of doctors, and improve surgical efficiency. This article is supported by the National Natural Science Foundation of China National Major Scientific Research Instrument Development Project "Highly Sensitive Intelligent Microscopic Precise Navigation Manipulator for Neurosurgical Tumor Resection" (62027813), to improve the intelligence of neurosurgery robot craniotomy, the problem of precise control in skull drilling and milling operations in robot-assisted craniotomy is investigated. The main content of the paper is as follows:

1. Aiming at the problem of how to make the robot autonomously perform the task of drilling the skull, a control method of feed force, feed velocity, and pose mixed constraints for the robot drilling was designed, and a craniotomy robot robot experimental platform was built to verify the control method.The craniotomy robot experimental platform built in this paper is equipped with a force perception module to simulate the sugeon's perception of force signal. In the robot autonomous skull drilling control method, the movement of the robot drilling is constrained to a straight line and the feed force is controlled by a bang-bang controller that limits the drilling velocity of the robotic arm when the feed force is too large. Through separated animals According to the feed force signal of the skull drilling experiment, the drilling state is classified into four phases: thrust force increasing phase, thrust force saturation phase, breakthrough phase and mechanical protection phase.

2. Aiming at the problem of how the robot can autonomously identify the skull penetration during the craniotomy drilling process, this paper designs a skull-meningeal boundary detection algorithm based on monitoring the change of the feed force signal. This method sets two continuous time series windows, the first continuous time series window T1 is used to track the change of feed force, and the mean value plays the role of filtering and reducing the influence of noise. The first continuous time series window T2 is used for buffering, reducing the impact of data entering T1 in the breakthrough phase on detection. Aiming at the problem of how to reasonably select the parameters of the skull-meningeal boundary detection algorithm, this paper established a linear approximation model for the relationship between the feed force and the penetration amount of the craniotomy drill in the breakthrough stage. The statistical data of the noise in the saturation stage is used as a constraint to transform the parameter selection of the cranial-meningeal boundary detection algorithm into an optimization problem, and the parameters of the cranial-meningeal boundary detection algorithm are set by solving the optimization problem. In this study, a total of 72 robotic craniotomy operations were used to conduct drilling experiments on the robot's autonomous drilling and autonomous detection of the skull-meningeal boundary. The experiments were carried out on the isolated heads of seven goat heads, one Pinger dog, and one Shiba Inu. . Among them, there were four cases where the experimental skull was not drilled through, but the mechanical protection device of the craniotomy drill was triggered in advance, and the experiment was forced to terminate; the experimental skull was drilled through in one case, but the surgical robot failed to recognize the signal; the surgical robot experimental platform could recognize it by itself. There were 67 experiments in which the skull was drilled through, accounting for 93.06%, and no intracranial soft tissue damage was observed in any of the experiments.

3. Aiming at the problem of robot autonomously milling skull, a force-position mixed control method based on PID fuzzy control is designed for robot milling skull. In this paper, the generation of the upper pull force is modeled as the elastic deformation of the "J"-shaped structure of the milling cutter protective sheath, and combined with the impedance model of the manipulator and the unknown environment, it is proposed that the control of the upper pull force of the manipulator loading the milling cutter can be controlled by using time-varying and position-dependent Linearly dependent PD controller for control. In order to deal with the problem of force-position mixed control with model uncertainty and unknown environment, this chapter uses fuzzy PID to control the pull-up force, and performs milling experiments on plastic skull models and isolated Shiba Inu skulls. The root mean square error of the upper pull force control in the plastic skull model experiment is 1.24N. The inner surface of the animal skull is not smooth, and there are many protrusions and ravines; in the real milling experiment of the isolated Shiba Inu skull, the Relatively complete biological anatomical structure, therefore, during the milling process of the isolated Shiba Inu skull, the dura mater has been in elastic contact with the milling cutter protective sheath, which makes the designed control method not sensitive to the milling cutter during the milling of the animal skull There are fluctuations in the control of the pulling force, but the pulling force is still maintained during the milling process, and the milling experiment did not damage the intracranial soft tissue of the isolated animal.