Effectiveness of Transfer Learning for Deep Learning-Based Electrocardiogram Analysis
4
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Objectives: Many deep learning-based predictive models evaluate the waveforms of electrocardiograms (ECGs). Becausedeep learning-based models are data-driven, large and labeled biosignal datasets are required. Most individual researchersfind it difficult to collect adequate training data. We suggest that transfer learning can be used to solve this problem and increasethe effectiveness of biosignal analysis. Methods: We applied the weights of a pretrained model to another model thatperformed a different task (i.e., transfer learning). We used 2,648,100 unlabeled 8.2-second-long samples of ECG II data topretrain a convolutional autoencoder (CAE) and employed the CAE to classify 12 ECG rhythms within a dataset, which had10,646 10-second-long 12-lead ECGs with 11 rhythm labels. We split the datasets into training and test datasets in an 8:2 ratio. To confirm that transfer learning was effective, we evaluated the performance of the classifier after the proposed transferlearning, random initialization, and two-dimensional transfer learning as the size of the training dataset was reduced. Allexperiments were repeated 10 times using a bootstrapping method. The CAE performance was evaluated by calculating themean squared errors (MSEs) and that of the ECG rhythm classifier by deriving F1-scores. Results: The MSE of the CAE was626.583. The mean F1-scores of the classifiers after bootstrapping of 100%, 50%, and 25% of the training dataset were 0.857,0.843, and 0.835, respectively, when the proposed transfer learning was applied and 0.843, 0.831, and 0.543, respectively, afterrandom initialization was applied. Conclusions: Transfer learning effectively overcomes the data shortages that can compromiseECG domain analysis by deep learning.
Abstract
Objectives: Many deep learning-based predictive models evaluate the waveforms of electrocardiograms (ECGs). Becausedeep learning-based models are data-driven, large and labeled biosignal datasets are required. Most individual researchersfind it difficult to collect adequate training data. We suggest that transfer learning can be used to solve this problem and increasethe effectiveness of biosignal analysis. Methods: We applied the weights of a pretrained model to another model thatperformed a different task (i.e., transfer learning). We used 2,648,100 unlabeled 8.2-second-long samples of ECG II data topretrain a convolutional autoencoder (CAE) and employed the CAE to classify 12 ECG rhythms within a dataset, which had10,646 10-second-long 12-lead ECGs with 11 rhythm labels. We split the datasets into training and test datasets in an 8:2 ratio. To confirm that transfer learning was effective, we evaluated the performance of the classifier after the proposed transferlearning, random initialization, and two-dimensional transfer learning as the size of the training dataset was reduced. Allexperiments were repeated 10 times using a bootstrapping method. The CAE performance was evaluated by calculating themean squared errors (MSEs) and that of the ECG rhythm classifier by deriving F1-scores. Results: The MSE of the CAE was626.583. The mean F1-scores of the classifiers after bootstrapping of 100%, 50%, and 25% of the training dataset were 0.857,0.843, and 0.835, respectively, when the proposed transfer learning was applied and 0.843, 0.831, and 0.543, respectively, afterrandom initialization was applied. Conclusions: Transfer learning effectively overcomes the data shortages that can compromiseECG domain analysis by deep learning.
참고문헌 (28)
- Attia ZI, "An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm : a retrospective analysis of outcome prediction", Lancet, 394, pp.861~867 (2019)
- van Ravenswaaij-Arts CM, "Heart rate variability", Ann Intern Med, 118, pp.436~447 (1993)
- Al-Zaiti S, "Machine learning-based prediction of acute coronary syndrome using only the pre-hospital 12-lead electrocardiogram", Nat Commun, 11, pp.3966~ (2020)
- Attia ZI, "Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram", Nat Med, 25, pp.70~74 (2019)
- Isakadze N, "How useful is the smartwatch ECG", Trends Cardiovasc Med, 30, pp.442~448 (2020)
- Pan SJ, "A survey on transfer learning", IEEE Trans Knowl Data Eng, 22, pp.1345~1359 (2009)
- Zhao W, "Research on the deep learning of the small sample data based on transfer learning", AIP Conf Proc, 1864, pp.020018~ (2017)
- "Transfusion:understanding transfer learning for medical imaging" (2019)[URL]
- "The utility of general domain transfer learning for medical language tasks" (2020)[URL]
- Van Steenkiste G, "Transfer learning in ECG classification from human to horse using a novel parallel neural network architecture", Sci Rep, 10, pp.186~ (2020)
- Tadesse GA, "Cardiovascular disease diagnosis using cross-domain transfer learning", Annu Int Conf IEEE Eng Med Biol Soc, 2019, pp.4262~4265 (2019)
- Salem M, "ECG arrhythmia classification using transfer learning from 2-dimensional deep CNN features", Proceedings of 2018 IEEE Biomedical Circuits and Systems Conference (BioCAS) (2018)
- Abdelazez M, "Transfer learning for detection of atrial fibrillation in deterministic compressive sensed ECG", Annu Int Conf IEEE Eng Med Biol Soc, 2020, pp.5398~5401 (2020)
- Wen L, "A new deep transfer learning based on sparse auto-encoder for fault diagnosis", IEEE Trans Syst Man Cybern Syst, 49, pp.136~144 (2017)
- Wang C, "Automated chest screening based on a hybrid model of transfer learning and convolutional sparse denoising autoencoder", Biomed Eng Online, 17, pp.63~ (2018)
- Eduardo A, "ECG-based biometrics using a deep autoencoder for feature learning-an empirical study on transferability", Proceedings of the 6th International Conference on Pattern Recognition Applications and Methods, pp.463~470 (2017)
- 윤덕용, "System for Collecting Biosignal Data from Multiple Patient Monitoring Systems", Healthcare Informatics Research, 23(4), pp.333~337 (2017), 대한의료정보학회[DOI]
- Zheng J, "A 12-lead electrocardiogram database for arrhythmia research covering more than 10, 000 patients", Sci Data, 7, pp.48~ (2020)
- Baldi, "Autoencoders, unsupervised learning, and deep architectures", Proceedings of the International Conference on Machine Learning (ICML): Unsupervised and Transfer Learning, pp.37~49 (2012)
- Liu W, "A survey of deep neural network architectures and their applications", Neurocomputing, 234, pp.11~26 (2017)
- "How convolutional neural network see the world-A survey of convolutional neural network visualization methods" (2018)[URL]
- Deng J, "ImageNet : a large-scale hierarchical image database", Proceedings of 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp.248~255 (2009)
- Szegedy C, "Going deeper with convolutions", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.1~9 (2015)
- DiCiccio TJ, "Bootstrap confidence intervals", Stat Sci, 11, pp.189~212 (1996)
- "Adam: a method for stochastic optimization" (2014)[URL]
- Hussain M, "Advances in computational intelligence systems", pp.191~202 (2018), Springer
- "What makes ImageNet good for transfer learning?" (2016)[URL]
- Porumb M, "Precision medicine and artificial intelligence : a pilot study on deep learning for hypoglycemic events detection based on ECG", Sci Rep, 10, pp.170~ (2020)