Wang Ludi, Zhou Wei, Liu Na, Xing Ying, Zhou Xiaoguang
Annu Int Conf IEEE Eng Med Biol Soc. 2018 Jul;2018:514-517. doi: 10.1109/EMBC.2018.8512300.
Heart rate variability has been proven to be an effective prediction of risk of heart failure. The tradition method required manual feature extraction, thus may lead to potential error. In order to improve the robustness, a deep learning method based on long short-term memory has been presented in this paper. Three RR interval length (N) for detection are used. Without pre-processing, this method obtain 82.47%, 85.13% and 84.91% accuracy for N=50 (average time length is 37. 8s), N=100 (average time length is 73. 9s), N=500 (average time length is 369. 5s), respectively. This method makes it possible to detect CHF through intelligent hardware or mobile application.
心率变异性已被证明是心力衰竭风险的有效预测指标。传统方法需要手动提取特征，因此可能会导致潜在误差。为了提高鲁棒性，本文提出了一种基于长短期记忆的深度学习方法。使用了三个用于检测的RR间期长度（N）。该方法无需预处理，对于N = 50（平均时长为37.8秒）、N = 100（平均时长为73.9秒）、N = 500（平均时长为369.5秒），分别获得了82.47%、85.13%和84.91%的准确率。该方法使得通过智能硬件或移动应用检测慢性心力衰竭成为可能。
