Analysis of ECG Signal by Using an FCN Network for Automatic Diagnosis of Obstructive Sleep Apnea.

Autor:	Ayashm, Sarah, Chehel Amirani, Mehdi, Valizadeh, Morteza
Předmět:	SLEEP apnea syndromes CONVOLUTIONAL neural networks ELECTROCARDIOGRAPHY DEEP learning RESPIRATORY obstructions
Zdroj:	Circuits, Systems & Signal Processing; Nov2022, Vol. 41 Issue 11, p6411-6426, 16p
Abstrakt:	Obstructive sleep apnea (OSA) is a sleep disorder in which a person's breathing is repeatedly interrupted because of airway obstruction. This study presents a fully convolutional neural network to diagnose the OSA by analyzing the single–lead ECG signal. This model does not employ any fully connected layers for classification, and the entire model only consists of 1D convolutional, activation, Pooling, and Batch Normalization layers. The main idea behind this model is to make decisions based on local segments of the signal, and the final classification is performed by aggregation of all decisions. This model implements the divide and conquer problem–solving method within a deep learning structure by the joint use of a convolutional layer and a global average pooling instead of a dense layer for classification. It was trained and evaluated by using a dataset available on the Physionet website. The accuracy of the proposed model outperformed other algorithms, which employ single lead ECG for the diagnosis of OSA. [ABSTRACT FROM AUTHOR]
Databáze:	Complementary Index
Externí odkaz:	Zobrazit plný text záznamu Full text from SpringerLink