A Multi-label Learning Method to Detect Arrhythmia Based on 12-Lead ECGs
Cardiovascular disease (CVD) is one of the most serious diseases that harm human life and gives a huge burden to the health care system. Recent advances in deep learning have achieved great success in object detection, speech and image recognition. Although deep learning has been applied to the detection of arrhythmia, detection accuracy is limited because of three major issues: 1. Each ECG signal maybe contains more than one-label information; 2. It is hard to classify ECG with different lengths; 3. Data imbalance problem is severe for arrhythmia detection. In this paper, we present a multi-label learning algorithm to address the class imbalance and detection on ECGs with different durations. We utilize Deep Convolutional Generative Adversarial Networks (DCGANs) and Wasserstein GAN-Gradient Penalty (WGAN-GP) to generate new positive samples and use two losses to balance the importance between positive samples and negative samples. Moreover, we construct a Squeeze and Excitation-ResNet (SE-ResNet) module for normal rhythm and arrhythmia detection. In order to solve the multi-label classification problem, we train nine different binary classifiers for each category and determine which types of rhythm the ECG signals belong to. Experimental results on The ECG Intelligence Challenge 2019 dataset demonstrate that our multi-label learning method achieves competitive performance in multi-label ECGs classification.
KeywordsArrhythmia ECG DCGANs WGAN-GP SE-ResNet Multi-label learning
This work was supported by the National Natural Science Foundation of China (No.61571628).
- 3.Ye, C., Kumar, B.V.K.V., Coimbra, M.T.: Combining general multi-class and specific two-class classifiers for improved customized ECG heartbeat classification. In: 2012 21st International Conference on Pattern Recognition (ICPR 2012). IEEE Computer Society (2012)Google Scholar
- 6.Moran, M.E., Soriano, M.C., Fischer, I., et al.: Electrocardiogram classification using reservoir computing with logistic regression. IEEE J. Biomed. Health Inform. 19(3) (2014)Google Scholar
- 7.Hong, S., Wu, M., et al.: ENCASE: an ensemble classifier for ECG classification using expert features and deep neural networks (2017)Google Scholar
- 12.Goodfellow, I.J., et al.: Generative adversarial nets. In: NIPS (2014)Google Scholar
- 13.Radford, A., Metz, L., Chintala, S.: Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv:1511.06434 (2015)
- 14.Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., Courville, A.C.: Improved training of wasserstein GANs. CoRR, abs/1704.00028 (2017)Google Scholar
- 15.Lin, T., Goyal, P., Girshick, R.B., He, K., Dollar, P.: Focal loss for dense object detection. In: IEEE ICCV (2017)Google Scholar
- 17.He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR (2016)Google Scholar
- 18.Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: CVPR (2018)Google Scholar
- 19.Kingma, D.P., Ba, J.L.: Adam: a method for stochastic optimization. In: Proceedings of International Conference on Learning Representations, pp. 1–15 (2015)Google Scholar