基于脉象分类的血压自适应连续测量

发布时间：2018-10-17 12:49

【摘要】：血压是衡量人体健康状况的重要指标,特别是连续血压,它能够间接反应出心脏和血管的运行状况,是临床上进行疾病诊断、治疗效果观察以及疾病预防判断的重要依据。但现在市面上的连续血压测量设备,主要为可穿戴式电子血压计都有着精确度差的缺点,不能准确的判断人体是否出现危险病态,所以对血压连续准确的测量以及异常状况的有效判断在预防心血管并发症以及对长期高血压患者的降压用药起到良好的监督作用和重要意义。因此,针对上述问题,本文对于血压连续测量提出了一种新型方法-基于脉象分类的血压自适应连续测量。该方法采用新型传感器一RF射频雷达实现对人体桡动脉脉搏波的双路信号获取,然后引入分层定势联想机制模型实现脉象的引导式自动分类,最后通过分级自适应血压预测模型实现血压的实时测量。论文的主要研究内容包括:(1)深入了解RF-射频雷达的工作原理以及其优势,设计一套人体桡动脉脉搏波采集系统,并利用Labview搭建出一套数据实时显示与数据保存系统。通过与脉搏波采集系统金标准进行对比验证了该系统的有效性。(2)脉象的准确分类是后期血压预测的基础,只有实现准确的分类才能保证血压预测的准确性。基于人类定势思维机制,我们提出了基于分层定势联想机制的脉象分类模型。首先进行友邦因子分析实现脉象粗分类并确定引导方向,然后利用定势联想神经网络实现对脉象的有效分类。神经元交互联想网络融合了引导式变异以及脉象演变规则具有较强的定势联想能力,可以有效地实现被测脉象与典型脉象的自联想。(3)在血压的预测阶段,我们引入了分级自适应血压预测模型。首先,通过脉象建立其与血压线性模型的内部联系,根据被测人的脉象以及相关信息实现一级血压模型的动态调整。然后,利用训练好的带参数库的PSO-BP神经网络实现二级血压最终结果的调整。实验结果表明,基于分层定势联想机制的脉象分类模型可以对人体常见脉象实现较高的分类准确度,其准确率达到92.86%,相比其他方法具有更好的分类效果。同时,血压自适应预测模型的预测准确度总体达到了 94.65%,能够对异常血压数据做到准确的判断,并且对个人血压的连续追踪趋势实现较好的跟随性。
[Abstract]:Blood pressure is an important index to measure human health, especially continuous blood pressure. It can indirectly reflect the operation of heart and blood vessels. It is an important basis for clinical diagnosis, observation of therapeutic effect and judgment of disease prevention. However, the existing continuous blood pressure measurement devices on the market, mainly wearable electronic sphygmomanometers, have the disadvantage of poor accuracy and cannot accurately judge whether the human body is in danger or not. Therefore, continuous and accurate measurement of blood pressure and effective judgment of abnormal condition play a good role in the prevention of cardiovascular complications and antihypertensive medication in long-term hypertensive patients. Therefore, to solve the above problems, this paper proposes a new method for continuous blood pressure measurement, which is adaptive continuous blood pressure measurement based on pulse classification. In this method, a new type of sensor, RF radio frequency radar, is used to obtain the dual signals of human radial pulse wave, and then the hierarchical setting association mechanism model is introduced to realize the guided automatic classification of pulse images. Finally, a hierarchical adaptive blood pressure prediction model is used to realize the real-time blood pressure measurement. The main contents of this paper are as follows: (1) deeply understand the working principle and advantages of RF- RF radar, design a pulse wave acquisition system of human radial artery, and set up a real-time data display and data storage system using Labview. The validity of the system is verified by comparing it with the gold standard of pulse wave acquisition system. (2) the accurate classification of pulse image is the basis of blood pressure prediction in the later stage, and the accuracy of blood pressure prediction can only be guaranteed by accurate classification. Based on the human fixed thinking mechanism, we propose a pulse classification model based on hierarchical stereotype association mechanism. Firstly, the coarse classification of pulse images is realized by friendly factor analysis and the guiding direction is determined. Then, the effective classification of pulse images is realized by using the fixed associative neural network. The neural interactive association network combines the guided mutation and pulse evolution rules, and has a strong ability of setting association, which can effectively realize the autoassociation between the measured pulse and the typical pulse. (3) in the stage of blood pressure prediction, We introduce a hierarchical adaptive blood pressure prediction model. Firstly, the internal relation between pulse and blood pressure linear model is established, and the first order blood pressure model is dynamically adjusted according to the pulse and related information. Then, a trained PSO-BP neural network with parameter library is used to adjust the final results of secondary blood pressure. The experimental results show that the classification model based on hierarchical fixed pattern association mechanism can achieve higher classification accuracy for common human pulse images, and the accuracy is 92.86, which is better than other methods. At the same time, the prediction accuracy of the adaptive blood pressure prediction model is 94.65, which can accurately judge the abnormal blood pressure data, and achieve a better follow-up to the continuous tracking trend of individual blood pressure.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R443.5

【参考文献】