Over-fitting suppression training strategies for deep learning-based atrial fibrillation detection

Zhang, Xiangyu, Li, Jianqing, Cai, Zhipeng, Zhang, Li, Chen, Zhenghua and Liu, Chengyu (2021) Over-fitting suppression training strategies for deep learning-based atrial fibrillation detection. Medical & Biological Engineering & Computing, 59 (1). pp. 165-173. ISSN 0140-0118

Different_training_strategy_for_deep_learning_based_AF_detection.pdf - Accepted Version

Download (423kB) | Preview
Official URL: https://doi.org/10.1007/s11517-020-02292-9


Nowadays, deep learning-based models have been widely developed for atrial fibrillation (AF) detection in electrocardiogram (ECG) signals. However, owing to the inevitable over-fitting problem, classification accuracy of the developed models severely differed when applying on the independent test datasets. This situation is more significant for AF detection from dynamic ECGs. In this study, we explored two potential training strategies to address the over-fitting problem in AF detection. The first one is to use the Fast Fourier transform (FFT) and Hanning-window-based filter to suppress the influence from individual difference. Another is to train the model on the wearable ECG data to improve the robustness of model. Wearable ECG data from 29 patients with arrhythmia were collected for at least 24 h. To verify the effectiveness of the training strategies, a Long Short-Term Memory (LSTM) and Convolution Neural Network (CNN)-based model was proposed and tested. We tested the model on the independent wearable ECG data set, as well as the MIT-BIH Atrial Fibrillation database and PhysioNet/Computing in Cardiology Challenge 2017 database. The model achieved 96.23%, 95.44%, and 95.28% accuracy rates on the three databases, respectively. Pertaining to the comparison of the accuracy rates on each training set, the accuracy of the model trained in conjunction with the proposed training strategies only reduced by 2%, while the accuracy of the model trained without the training strategies decreased by approximately 15%. Therefore, the proposed training strategies serve as effective mechanisms for devising a robust AF detector and significantly enhanced the detection accuracy rates of the resulting deep networks.

Item Type: Article
Uncontrolled Keywords: Atrial fibrillation (AF), Electrocardiogram (ECG), Deep learning, Wearable ECG
Subjects: B800 Medical Technology
B900 Others in Subjects allied to Medicine
G400 Computer Science
Department: Faculties > Engineering and Environment > Computer and Information Sciences
Depositing User: John Coen
Date Deposited: 20 Jan 2021 11:35
Last Modified: 02 Jan 2022 03:30
URI: http://nrl.northumbria.ac.uk/id/eprint/45272

Actions (login required)

View Item View Item


Downloads per month over past year

View more statistics