基于空洞卷积神经网络的心律失常分类算法

本文提出了一种基于卷积网络的心电信号分类算法，设计了空洞卷积池化金字塔模块，通过不同尺寸的空洞卷积提取信息，再将各通道的信息聚合，在增强网络的特征提取能力的同时可以降低参数量。本文聚焦于窦性心律、房性早搏、心动过速以及心动过缓4种分类，使用的心电图数据集来自医院的实测数据，数据集包含75000名不同检测者的心电记录。经过测试，本文提出的模型在该数据集上取得了0.89的F1值，另外在CinC2017数据集上也达到了0.87的F1值。实验结果表明该分类算法具有优秀的特征提取和分类能力，在心电信号的实时分类中具备应用前景。

Arrhythmia detection algorithm based on dilated convolutional neural network

Abstract: An electrocardiogram (ECG) signal classification algorithm based on convolutional network is proposed. The algorithm adopts an atrous spatial pooling pyramid module to extract information through atrous convolution of different sizes, and aggregates the information of each channel for enhancing the ability of feature extraction and reducing the number of parameters. The study focuses on the categories of sinus rhythm, premature atrial contraction, tachycardia and bradycardia, and a real ECG data set from a hospital which contains ECG records of 75 000 different subjects is used for experiment. The results reveal that the proposed model reaches an F1 score of 0.89 on the real ECG data set, and also achieved an F1 score of 0.87 on the CinC2017 data set, which indicates that the classification algorithm has excellent feature extraction and classification capabilities, and has application prospects in the real-time classification of ECG signals.