深度学习技术对心电图进行自动的疾病诊断具有十分重要的意义.但现有的分类算法存在计算速度慢、实时性差以及对心电信号多尺度特征利用不充分的问题,会对某些疾病产生漏检,影响自动诊断技术的效率和精确度.因此提出了一种融合注意力机制与多尺度特征提取的轻量化心电图多标签分类网络(Lightweight Network with Attention for Multi Scale Classification,LAMSCN).该模型可以有效地识别多种心脏病症状.实验结果表明,与MobileNet等主流模型相比,LAMSCN有效降低了模型参数量,同时对17种疾病的分类性能指标F1可以达到0.905,极大降低了对部署设备的要求.
A lightweight network incorporating attention mechanisms used for multilabel classification of the electrocardiogram
The use of deep learning technology is of great significance for the automatic diagnosis of diseases on electro⁃cardiogram(ECG).However,the existing classification algorithms have the problems of slow calculation speed,poor real-time performance,and insufficient utilization of multi-scale characteristics of ECG signals.It will result in missed detec⁃tion of certain diseases and affect the efficiency and accuracy of automatic diagnosis technology.Therefore,a lightweight network with attention for multi scale classification(LAMSCN)is presented.The model can effectively identify patients with multiple heart conditions at the same time.Experimental results show that compared with mainstream models such as MobileNet,LAMSCN effectively reduces the amount of model parameters and the classification performance index F1 of 17 diseases can reach 0.905,which greatly reduces the requirements for deployed equipment.
万方数据
维普
