摘要
心电图(ECG)是一种常规的身体监测手段,通过分析人体在不同状态下心电活动的变化,评估其心血管健康状况。考虑到人体生理特征的差异,以及不同状态下心电活动规律的变化,如何设计一种自动适应各种场景的心电信号分类模型具有重要的现实意义。该文创新性地将心电信号转化为图像数据,并采用可微分神经网络架构搜索算法(PC-DARTS)对不同分布的心电检测数据自动搭建最优神经网络模型,实现了不同场景下心电信号的精准分类。分别在心律失常数据集PhysioNet MIT-BIH和诊断性心电图数据集PTB上进行心电信号分类实验,以验证所提方法在不同应用场景下的辨识性能。实验结果表明,与其他方法相比,该文算法具备更高的准确度和更强的鲁棒性,同时,能够应对不同采集设备、实验环境以及被试人群所带来的分类辨识挑战,具备较强的泛化性能。未来,该研究成果有望与新型心电监测设备相结合,实现高效精准的心电检测功能,加速心电检测在更多领域中的落地与应用。
Electrocardiogram(ECG)is a commonly used method for monitoring the body.The analysis of alterations in electrocardiographic activity during physical exercise can evaluate an individual’s cardiovascular health status.Designing an automatic adaptive ECG signal classification model that can accommodate different physiological characteristics and changes in ECG activity patterns in different states is of great practical significance.In this study,we proposed an innovative approach that transformed ECG signals into image data and automatically constructed optimal neural network models for different distributions of ECG detection data using the differentiable neural architecture search algorithm(PC-DARTS).This approach achieved accurate ECG signal classification in various scenarios.In this paper,ECG signal classification experiments were conducted on the PhysioNet MIT-BIH arrhythmia dataset and the PTB diagnostic ECG dataset,respectively,to verify the discrimination performance of the proposed method in different application scenarios.The results indicate that the algorithm proposed in this paper has higher accuracy and greater robustness compared to other methods.At the same time,the proposed method is able to cope with the classification challenges posed by different acquisition devices,different experimental environments and different subject populations,and has sufficient generalization performance.In the future,the research results are expected to be combined with new ECG monitoring devices to achieve efficient and accurate ECG detection and accelerate the application of ECG detection in more fields.
作者
宋群
袁青霞
王俊江
SONG Qun;YUAN Qingxia;WANG Junjiang(School of Automation,Northwestern Polytechnical University,Xi’an,710072,China;Zhongke Ruiyan(Tianjin)Technology Co.,Ltd.,Tianjin 300350,China;Department of Physical Education,Tianjin Chengjian University,Tianjin 300384,China)
出处
《西北大学学报（自然科学版）》
CAS
CSCD
北大核心
2023年第5期771-781,共11页
Journal of Northwest University（Natural Science Edition）
基金
国家自然科学基金(62006192)。
关键词
心电检测
运动健康
神经网络
自动机器学习
神经架构搜索
electrocardiogram
sports health
neural networks
automated machine learning
neural architecture search
