PerDetect: A Personalized Arrhythmia Detection System Based on Unsupervised Autoencoder

Cardiovascular disease has become a common cause of death. Arrhythmia is a common cardiovascular disease. Cardiovascular disease has become a common cause of death. Arrhythmia is a common cardiovascular disease. Deep learning has been widely used in arrhythmia detection. However, the application of...

Full description

Saved in:

Bibliographic Details
Published in:	2023 7th Asian Conference on Artificial Intelligence Technology (ACAIT) pp. 914 - 919
Main Authors:	Zhong, Zhaoyi, Sun, Le
Format:	Conference Proceeding
Language:	English
Published:	IEEE 10.11.2023
Subjects:	Arrhythmia Arrhythmia detection autoencoder ECG signal Electrocardiography Heart beat Real-time systems Training Training data unsupervised Learning
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Cardiovascular disease has become a common cause of death. Arrhythmia is a common cardiovascular disease. Cardiovascular disease has become a common cause of death. Arrhythmia is a common cardiovascular disease. Deep learning has been widely used in arrhythmia detection. However, the application of unsupervised learning to arrhythmia detection is not extensive. This paper proposes a personalized arrhythmia detection system PerDetect based on an unsupervised autoencoder. The system trains a separate BiLSTM-based autoencoder BiAE for each patient for arrhythmia detection. BiAE only needs to use the patient's normal heartbeat for training, which effectively avoids the problem of data imbalance. We carried out experiments on MIT-BIH Arrhythmia Database. Experiments show that the system only needs a small amount of ECG training data (within five minutes) to achieve good performance. The ACC of our method on MIT-BIH Arrhythmia Database is 97%.
DOI:	10.1109/ACAIT60137.2023.10528650