Optimized Convolutional Neural Network by Genetic Algorithm for the Classification of Complex Arrhythmia

被引：3

作者：

Qian, Li ^{[1
]}

Wang, Jianfei ^{[1
]}

Jin, Lian ^{[1
]}

Huang, Yanqi ^{[1
]}

Zhang, Jiayu ^{[1
]}

Zhu, Honglei ^{[1
]}

Yen, Shengjie ^{[1
]}

Wu, Xiaomei ^{[1
,2
,3
]}

机构：

[1] Fudan Univ, Elect Engn Dept, Shanghai 200433, Peoples R China

[2] Shanghai Key Lab Med Image Comp & Comp Assisted I, Shanghai 200032, Peoples R China

[3] Shanghai Engn Res Ctr Assist Devices, Shanghai 200433, Peoples R China

来源：

JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS | 2019年 / 9卷 / 09期

基金：

中国国家自然科学基金;

关键词：

Multiclassification; Complex Arrhythmia; Convolutional Neural Network; Electrocardiogram; Genetic Algorithm; Grid Search; SUPPORT VECTOR MACHINES; ATRIAL-FIBRILLATION; PATTERN-RECOGNITION; SIGNALS; MODEL;

D O I：

10.1166/jmihi.2019.2813

中图分类号：

Q [生物科学];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

Complex arrhythmia is a serious cardiovascular disorder. Delayed diagnosis of complex arrhythmia can cause serious consequences, such as blood clots, stroke, heart failure, and sudden death. However, for accurate diagnosis and treatment, correct identification of the type of arrhythmia is important. Therefore, in this study, for automatic multiclassification of surface electrocardiogram signals (i.e., atrial fibrillation [AFIB], atrial flutter [AFL], supraventricular tachycardia [SVTA], ventricular tachycardia [VT], ventricular flutter [VFL], sinus rhythm [N], and NOISE), an optimized 17-layer-deep convolutional neural network (CNN) was constructed using a genetic algorithm and grid search. Various, indicators were applied to evaluate the classification performance of the constructed CNN. The resulting accuracy, sensitivity, specificity, precision, and macro_F1 score of the constructed CNN were 95.87%, 95.78%, 99.18%, 96.02%, and 95.87%, respectively. The results showed that the proposed CNN can be a powerful tool for multiclassification of complex arrhythmia.

引用

页码：1905 / 1912

页数：8