作者：唐塬, 常俊杰, 彭少雄, 曾辉, 魏麟权

作者单位：重庆邮电大学光电工程学院，重庆 400065

关键词：栓子检测;相关性分析;EMD;神经网络

摘要：

栓子的准确检测能为早期脑血管疾病诊断提供可靠的依据，不同大小的微栓子对脑血管造成的损伤不同。针对微栓子大小问题，使用三种直径大小分别为0.08、0.15 、0.25 mm的尼龙颗粒模拟血液中的栓子，并提出经验模态分解(empirical mode decomposition, EMD)结合BP神经网络模式识别的超声栓子信号分类方法。首先对不同直径大小的微栓子回波信号进行EMD处理，得到各阶的固有模函数(intrinsic mode function, IMF)，再分析IMF分量和栓子回波信号的相关性，将相关性较高的前五阶IMF分量在时域和频域的特征参数共20个输入BP神经网络进行识别分类，其中神经网络输入节点对应特征值，输出节点对应不同直径大小的微栓子。实验结果表明该方法能准确对三种不同直径大小微栓子进行识别与分类，且识别率达到98.3%。

Research on ultrasonic emboli signal classification based on EMD and neural network
TANG Yuan, CHANG Junjie, PENG Shaoxiong, ZENG Hui, WEI Linquan
School of Optoelectronic Engineering, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
Abstract: Accurate detection of emboli can provide a reliable basis for the early diagnosis of cerebrovascular diseases. Different sizes of microemboli cause different damage to cerebral blood vessels. Aiming at the size of micro-emboli, three nylon particles with diameters of 0.08 mm, 0.15 mm, and 0.25 mm were used to simulate emboli in blood, and an empirical mode decomposition (EMD) combined with BP neural network was proposed. Ultrasound embolic signal classification method for pattern recognition. Firstly, EMD processing was performed on the echo signals of microemboli with different diameters, and the intrinsic mode function (IMF) of each order was obtained. A total of 20 eigenparameters of the high first-fifth-order IMF components in the time domain and frequency domain are input to the BP neural network for identification and classification. The input nodes of the neural network correspond to the eigenvalues, and the output nodes correspond to microemboli of different diameters.The experimental results show that the method can accurately identify and classify three kinds of microemboli with different diameters, and the recognition rate reaches 98.3%.
Keywords: embolism detection;correlation analysis;EMD;neural networks
2023, 49(8):15-20  收稿日期: 2022-03-21;收到修改稿日期: 2022-06-25
基金项目: 国家自然科学基金（11464030）
作者简介: 唐塬(1996-),男,重庆市人,硕士研究生,专业方向为超声无损检测技术
参考文献
[1] MOLLOY J, MARKUS H S. Asymptomatic embolization predicts stroke and TIA risk in patients with carotid artery stenosis[J]. Stroke, 1999, 30(7): 1440-1443
[2] 赵杨, 沈卫平, 郭志力, 等. 微栓子成分和大小与脑组织损害相关性的实验研究[C]//第九次全国神经病学学术大会论文汇编, 2006: 100-101.
[3] PATTERSON R H, KESSLER J. Microemboli during cardiopulmonary bypass detected by ultrasound[J]. Surgery Gynecology & Obstetrics, 1969, 129(3): 505
[4] MARKUS H S, TEGELER C H. Experimental aspects of high-intensity transient signals in the detection of emboli[J]. Journal of Clinical Ultrasound, 1995, 23(2): 81-87
[5] 陈曦, 汪源源, 王威琪. 基于小波包的血栓超声多普勒信号检测[J]. 声学学报, 2004, 29(4): 341-345
[6] 陈一骄, 汪源源, 周华琳, 等. 基于自适应小波包基的超声多普勒栓子信号检测[C]//中国生物医学工程进展--2007中国生物医学工程联合学术年会论文集(上册), 2007: 671-675.
[7] SERBES G, SAKAR B E, GULCUR H O, et al. An emboli detection system based on Dual Tree Complex Wavelet Transform and ensemble learning[J]. Applied Soft Computing, 2015(37): 87-94
[8] BENOUDJIT N, FERROUDJI K, BAHAZ M, et al. In vitro microemboli classification using neural network models and RF signals[J]. Ultrasonics, 2011, 51(3): 247-252
[9] SOMBUNE P, PHIENPHANICH P, PHUECHPANPAISAL S, et al. Automated embolic signal detection using deep convolutional neural network [C]//The 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’17), 2017: 3365-3368.
[10] PIERLEONI P, PALMA L, BELLI A, et al. An EMD-based algorithm for emboli detection in echo doppler audio signals[J]. Electronics, 2019, 8(8): 824
[11] 陈志恒, 罗文斌, 常俊杰, 等. 基于EMD的神经网络空耦超声储油罐液位检测[J]. 中国测试, 2021, 47(1): 9-14
[12] 徐修萍. 超声多普勒人工心脏血栓检测系统的研究[D]. 上海: 上海交通大学, 2010.
[13] HUANG N E, SHEN Z, LONG S R, et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis[J]. Proceedings Mathematical Physical & Engineering Sciences, 1998, 454(1971): 903-995
[14] 张琦, 吕晓丽, 谢建华. 彩色速度成像及定量技术对正常成人颈动脉血流动力学的检测[J]. 第二军医大学学报, 2001, 22(3): 236-238
[15] 徐锋, 刘云飞. 基于EMD的胶合板损伤声发射信号特征提取及神经网络模式识别[J]. 振动与冲击, 2012, 31(15): 30-35
[16] 刘通, 秦正山, 何芳, 等. 基于神经网络的磁性活性炭吸附双酚A研究[J]. 磁性材料及器件, 2019, 50(1): 9-11