【作者】 沈鹏；

【导师】 高志强； 姜鸿； 吕威；

【作者基本信息】 中国协和医科大学 ， 耳鼻咽喉头颈外科学， 2010， 博士

【摘要】 研究背景和目的耳科手术中高速旋转的电钻钻头很容易碰到正常组织而造成副损伤,如何减少甚至避免电钻造成的副损伤,一直是有待解决的问题。一些学者尝试用导航或手术机器人的方法来控制电钻的运行,目前大多处于实验室阶段。由实际耳科手术经验可知,当术中出现钻头打滑、磨穿骨质、棉片缠绕等异常情况时容易造成正常组织结构的损伤。在对钻头受力分析的基础上,结合多传感器信息融合、微弱信号检测等相关技术,我们对正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕等各种情况下钻头受力的变化情况进行分析,探索其规律,并利用该规律对电钻的运行状态进行实时识别,最终实现耳科手术电钻的智能化控制。近年来,多传感器信息融合技术和微弱信号检测技术的快速发展和广泛应用为本课题的成功实施提供了较强的技术背景。材料与方法基于耳科手术电钻运行时钻头的受力分析,首先对第一台耳科手术电钻进行改装并合理布置了多维力传感器、电流传感器、电压传感器和转速传感器,采用单一条件(同一操作者、最大电压和3mm切割钻头),在新鲜猪肩胛骨上模拟进行正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕等四种情况各1000次,采集传感器信号,分析其规律性,并提取出各种异常情况下的特征信号；设计合适的多传感器信息融合系统识别电钻的运行状态,统计识别率,并设计停机程序使电钻在出现异常运行状态时自动快速地停止磨削。然后对第二台耳科手术电钻进行改装并安装了多维力传感器、电流传感器和电压传感器,采用复杂条件(最大电压,10名不同的操作者,直径2.5mm、4mm、5.9mm的切割钻头和直径4.2mm的金刚石钻头)下,在尸头颞骨上,每名操作者应用每种钻头模拟进行正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕等四种情况各100次,采集传感器信号进行分析,应用多传感器信息融合系统识别电钻的运行状态,统计识别率。结果单一条件下,在正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕等每种情况下,各传感器信号变化的趋势和过程大部分是一致的,可重复性强,有较强的规律性,能够提取出异常情况(包括钻头打滑、磨穿骨质和棉片缠绕)发生初期的特征信号；运用我们设计的BP神经网络,对正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕的识别率分别为82.2%、75.6%、71.6%和70.2%；当融合系统识别出电钻出现异常情况时,自动停机程序反应快速准确,能在0.2-0.3秒之内使电钻停止磨削。复杂条件下,各传感器信号变化的趋势和过程大部分一致,可重复性较强,有较强的规律性,与单一条件下的结果相似；我们设计的BP神经网络对复杂情况下正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕的平均识别率分别为81.3%、72.625%、68.575%和70.5%。结论本实验利用多种传感器对各种情况下钻头受力变化过程进行分析,发现在正常磨削(包括正常离开)、钻头打滑、磨穿骨质、棉片缠绕等四种情况发生过程中,钻头的受力变化有一定的规律性,能够提取出三种异常情况发生时的特征信号,经合适的BP神经网络融合后能较好的识别电钻的运行状态,为最终实现耳科手术电钻的智能化控制提供了较好的基础。但还需要进一步的研究,这部分工作我们正在进行中。更多还原

【Abstract】 BackgroundIn otologic surgery, the high-speed rotating drill bit is easy to touch important healthy structures and cause collateral damage, so minimizing or avoiding damage caused by loss of control during drilling is an important issue. Thus far, some navigational or robotic concepts for guiding the drill have been pursued in experimental temporal bone surgery, but certain key problems have not been resolved, and these studies have not yet led to clinical application.Practical experience of otologic surgery indicates that damage to important healthy structures is most likely to occur during drilling faults, such as drill bit slippage, drilling through the bone tissue wall and cotton swab entanglement. Using multi-sensor information fusion and weak signal detection techniques, we want to explore the change rule of forces acting on the drill bit during normal drilling (including normal removal of the drill bit from the working surface), drill bit slippage, drilling through the bone tissue wall and cotton swab entanglement, and use the rule to identify the drilling faults in real time and make an intelligent otologic drill finally.The rapid development and wide application of multi-sensor information fusion and weak signal detection techniques ensure the success of this research.Materials and MethodsBased on the analysis of forces acting on the drill bit when drilling, the first otologic drill was modified and equipped with force sensor, current sensor, voltage sensor and speed sensor. Under consistent conditions (the same surgeon, maximum voltage and a 3mm diameter cutting burr), the modified drill was used to simulate four different drilling scenarios in otologic surgery, including normal drilling (including normal removal of the drill bit from the working surface), drill bit slippage, drilling through the bone tissue wall and cotton swab entanglement, each scenario was repeated about 1000 times on fresh porcine scapulas. During the trials all sensor signals were recorded and analyzed the rule, then the significant sudden changes in the signals were extracted as characteristic signals. A multi-sensor information fusion system was designed to identify the drilling faults and to add up the identification rate, and a stop program was designed to make the drill stop drilling when the drilling faults were identified.Then the second otologic drill was modified and equipped with force sensor, current sensor and voltage sensor. Under different conditions(the maximum voltage,10 different otologic doctors, cutting burr of 2.5mm、4mm、5.9mm diameter and a 4.2mm diameter diamond bit), each doctor used each kind of drill bit to simulate each scenario for 100 times on the cadaveric temporal bone, all sensor signals were recorded and analyzed. The information fusion system was used to identify the drilling faults and to add up the identification rate.ResultsUnder consistent conditions, the signal of each sensor changed consistently during each drilling scenario, with high repeatability and regularity of signal variation. It is possible to extract a characteristic signal for each kind of drilling fault. Using our multi-sensor information fusion system(BP neural networks), the rate of identifying normal drilling (including normal removal of the drill bit from the working surface) was 82.2%, drill bit slippage was 75.6%, drilling through the bone tissue wall was 71.6% and cotton swab entanglement was 70.2%. The stop program made the drill stop drilling in 0.2-0.3 seconds when the drilling faults was identified.Under different conditions, the signal of each sensor changed consistently too, with high repeatability and regularity of signal variation,like the results of consistent conditions. The average identification rate was 81.3%,72.625%,68.575% and 70.5% respectively.ConclusionsThis study shows that, during normal drilling (including normal removal of the drill bit from the working surface), drill bit slippage, drilling through the bone tissue wall and cotton swab entanglement during otologic surgery, the forces acting on the drill bit change predictably under different conditions, characteristic signals can be extracted from three kinds of drilling faults. Using suitable BP neural networks, the drilling faults can be identified. This provides a good foundation of predicting the drilling faults and controlling the drill automatically. Further experiments are necessary to be done, these are underway.更多还原