| 650 | 7 | 193 |
| 下载次数 | 被引频次 | 阅读次数 |
心血管疾病是导致人类死亡的主要原因之一，在心血管疾病的诊断上，传统的医学方法需要有大量知识储备的专业人员，在高素质从医人员较少的欠发达地区易出现误诊的现象。为了节约医疗资源，降低心血管疾病误诊率，基于历史医疗数据，提出一种基于经验模态分解（empirical mode decomposition,EMD）与人工神经网络（artificial neural network,ANN）的新型面向心血管疾病数据分类模型，在分类模型中引入注意力机制（SE-Net模块），提高了检测效率。将所提出模型试用于多个公开数据集上，在MIT-BIH心律不齐数据库、欧洲ST-T心电数据库上精准率均表现良好。实验结果表明：在数据去噪任务中，文章提出模型的去噪性能明显优于基准模型；在疾病分类任务中，所提出模型的精准率达到95.78%，召回率达到86.16%，优于目前主流分类模型。同时，还设计了消融实验，用于验证两个模块的必要性。消融实验结果显示，文中两个模块对分类的准确性和鲁棒性均有提升，具有充分的必要性。
Abstract：Cardiovascular disease is one of the leading causes of global mortality. Traditional medical methods for cardiovascular disease diagnosis require a large amount of knowledge and expertise, leading to a high risk of misdiagnosis in underdeveloped regions with limited access to skilled medical professionals. To save medical resources and reduce the misdiagnosis rate of cardiovascular disease, a novel classification model based on empirical mode decomposition(EMD) and artificial neural networks(ANN) is proposed, leveraging historical medical data. The model incorporates the SE-Net module, which introduces an attention mechanism to enhance detection efficiency. The proposed model has been applied to multiple publicly available datasets, demonstrating excellent accuracy in both the MIT-BIH arrhythmia database and the European ST-T electro cardio gram(ECG) database. Experimental results indicate that the denoising performance of the proposed model surpasses that of the baseline model in noise reduction tasks. Furthermore, in disease classification tasks, the proposed model achieves a precision rate of 95.78%, and a recall rate of86.16%, outperforming current mainstream classification models. Additionally, ablation experiments have been designed to validate the necessity of the two modules. The results of the ablation experiments show that both modules contribute to the accuracy and robustness of the classification, highlighting their essentiality.
[1]李鸿强，吴非凡，曹路，等.基于改进深度残差网络的心电信号分类算法[J].天津工业大学学报，2022, 41(5):65-72.LI H Q, WU F F, CAO L, et al. ECG signal classification algorithm based on improved depth residual network[J].Journal of Tiangong University, 2022, 41(5):65-72.(in Chinese)
[2] SUN F P, NI Y, LUO Y H, et al. ECG classification based on Wasserstein scalar curvature[J]. Entropy, 2022,24(10):1450.
[3] ELIAS P, POTERUCHA T J, RAJARAM V, et al. Deep learning electrocardiographic analysis for detection of leftsided valvular heart disease[J]. Journal of the American College of Cardiology, 2022, 80(6):613-626.
[4] SHUKLA N, PANDEY A, SHUKLA A P, et al. ECGViT:a transformer-based ECG classifier for energy-constraint wearable devices[J]. Journal of Sensors, 2022, 2022:2449956.
[5] SHARMA K, ESKICIOGLU R. Deep learning-based ECG classification on raspberry PI using a tensorflow lite model based on PTB-XL dataset[J]. International Journal of Artificial Intelligence and Applications, 2022, 13(4):55-66.
[6]左立洋，刘明，杨畅.基于半监督卷积神经网络的心律不齐分类[J].激光杂志，2021, 42(12):219-225.ZUO L Y, LIU M, YANG C. Arrhythmia classification based on semi-supervised convolutional neural network[J].Laser Journal, 2021, 42(12):219-225.(in Chinese)
[7]袁高腾，周晓峰，郭宏乐.基于特征选择算法的ECG信号分类[J].山东大学学报（工学版），2022, 52(4):38-44.YUAN G T, ZHOU X F, GUO H L. ECG signal classification based on feature selection algorithm[J]. Journal of Shandong University(Engineering Science), 2022, 52(4):38-44.(in Chinese)
[8]郭树言.基于IEMD的ECG去噪和QRS提取算法研究[D].太原：中北大学，2021.GUO S Y. Research on ECG denoising and QRS detection algorithm based on IEMD[D]. Taiyuan:North University of China, 2021.(in Chinese)
[9]崔建伟，王冬青，刘金燕.基于均值滤波和小波变换的单幅图像去雾算法[J].计算机与数字工程，2022, 50(6):1339-1342.CUI J W, WANG D Q, LIU J Y. Single image dehazing algorithm based on average filtering and wavelet transform[J]. Computer and Digital Engineering, 2022, 50(6):1339-1342.(in Chinese)
[10]玛利亚木古丽·麦麦提，吐尔洪江·阿布都克力木，阿卜杜如苏力·奥斯曼，等.结合小波变换和同态滤波的医学图像增强算法[J].电子设计工程，2020, 28(24):1-5.MAMAITI M, ABDUKIRIM T, AOSIMAN A, et al. Medical image enhancement algorithm combining wavelet transform and homomorphic filtering[J]. Electronic Design Engineering, 2020, 28(24):1-5.(in Chinese)
[11]张丽丽，张伟.基于小波变换的滤波方法的研究与应用[J].数字技术与应用，2020, 38(10):25-26.ZHANG L L, ZHANG W. Research and application of filtering method based on wavelet transformation[J]. Digital Technology and Application, 2020, 38(10):25-26.(in Chinese)
[12]周卫佳.基于EMD和MP方法的医学信号和图像分析[D].北京：北京邮电大学，2010.ZHOU W J. The analyze of medical signal and imaging based on EMD and MP method[D]. Beijing:Beijing University of Posts and Telecommunications, 2010.(in Chinese)
[13]杨朋威，万全，刘春晖，等.基于EMD滤波改进的直流电流互感器现场测试研究[J].电气传动，2021, 51(13):8-13.YANG P W, WAN Q, LIU C H, et al. DC current transformer field test system based on EMD filtering[J]. Electric Drive, 2021, 51(13):8-13.(in Chinese)
[14]岳廷明，王金海，王慧泉.融合加速度信息的动态心电EMD滤波算法[J].计算机工程与应用，2015, 51(20):192-197.YUE T M, WANG J H, WANG H Q. Integration of accelerometer information and EMD filtering algorithm in dynamic ECG acquisition[J]. Computer Engineering and Applications, 2015, 51(20):192-197.(in Chinese)
[15]彭云.基于ICA和EMD的睡眠脑电图预处理算法的研究[D].重庆：重庆大学，2012.PENG Y. Study on pre-processing algorithms for sleep EEG based on Ica and EMD[D]. Chongqing:Chongqing University, 2012.(in Chinese)
[16]王野，李钦奉，齐继阳.基于改进自适应中值滤波的图像降噪方法[J].舰船电子工程，2022, 42(6):112-115.WANG Y, LI Q F, QI J Y. Image denoising method based on improved adaptive Median filter[J]. Ship Electronic Engineering, 2022, 42(6):112-115.(in Chinese)
[17]张雪峰，闫慧.基于中值滤波和分数阶滤波的图像去噪与增强算法[J].东北大学学报（自然科学版），2020, 41(4):482-487.ZHANG X F, YAN H. Image denoising and enhancement algorithm based on Median filtering and fractional order filtering[J]. Journal of Northeastern University(Natural Science), 2020, 41(4):482-487.(in Chinese)
[18]李兴秀.基于深度学习的心电信号心律失常分类方法研究[D].南昌：江西农业大学，2022.LI X X. Research on ECG arrhythmia classification method based on deep learning[D]. Nanchang:Jiangxi Agricultural University, 2022.(in Chinese)
[19]秦博.基于深度学习的多导联心电图分类识别算法研究[D].黄石：湖北师范大学，2022.QIN B. Multi-lead ECG signal recognition algorithm based on deep learning[D]. Huangshi:Hubei Normal University,2022.(in Chinese)
[20]刘伟洁.基于深度学习的ECG心律失常分类研究[D].西宁：青海师范大学，2022.LIU W J. Research on ECG arrhythmia classification model based on deep learning[D]. Xining:Qinghai Normal University, 2022.(in Chinese)
[21]刘守华，王小松，刘昱.基于深度学习的临床心电图分类算法[J].计算机与现代化，2021(8):52-57.LIU S H, WANG X S, LIU Y. Clinical electrocardiogram classification algorithm based on deep learning[J]. Computer and Modernization, 2021(8):52-57.(in Chinese)
[22]马晶，李林献，邱筱岷，等.基于卷积神经网络的心电图分类研究[J].医疗卫生装备，2020, 41(12):31-34.MA J, LI L X, QIU X M, et al. Research on ECG classification based on convolutional neural network[J]. Chinese Medical Equipment Journal, 2020, 41(12):31-34.(in Chinese)
[23] HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8):2011-2023.
[24]潘从成.基于深度学习和注意力机制的心电信号分类方法研究[D].南昌：江西财经大学，2022.PAN C C. Research on ECG signal classification method based on deep learning and attention mechanism[D]. Nanchang:Jiangxi University of Finance and Economics, 2022.(in Chinese)
[25]崔凯兴.面向时间序列的ECG诊断算法的设计与实现[D].北京：中国科学院大学（中国科学院沈阳计算技术研究所），2022.CUI K X. Design and implementation of time series-oriented ECG diagnosis algorithm[D]. Shenyang:Institute of Computing Technology, Chinese Academy of Sciences,2022.(in Chinese)
基本信息:
DOI：10.12194/j.ntu.20221205001
中图分类号:R540.41;TN911.7;TP18
引用信息:
[1]高轩瑞,曹进德,张金壬.一种基于深度学习的心电图分类方法[J],2023,22(03):17-25.DOI:10.12194/j.ntu.20221205001.
基金信息: