Correo Electrónico
DNINFOA - SIA
Bibliotecas
Convocatorias
Identidad U.N.
Published
Lightweight Deep Learning for Atrial Fibrillation Detection: Efficient Models for Wearable Devices
Aprendizaje profundo ligero para la detección de fibrilación auricular: modelos eficientes para dispositivos portátiles
DOI:
https://doi.org/10.15446/ing.investig.114530Keywords:
cardiac arrhythmia, deep learning, detection, ECG, electrocardiogram, machine learning, portable device (en)arritmia cardiaca, aprendizaje profundo, detección, ECG, electrocardiograma, aprendizaje automático, dispositivo portátil (es)
Downloads
The timely and accurate detection of atrial fibrillation (AF) is crucial for an early intervention and for reducing the associated health risks. Wearable technology has emerged as a viable solution for continuous AF monitoring, but deploying accurate AF detection models on resource-constrained devices remains a challenge due to the high computational and memory demands. This study proposes a lightweight and efficient deep learning approach for real-time AF diagnosis in portable devices. We designed a series of convolutional neural network (CNN) models optimized for high accuracy while maintaining a minimal computational footprint. To further enhance efficiency, we explored deep learning compression techniques, including pruning, quantization, and knowledge distillation. Our results demonstrate that the proposed models achieve state-of-the-art accuracy while significantly reducing memory usage and computational complexity, making them suitable for real-time deployment. Additionally, we validated their feasibility by implementing them on a microcontroller, showcasing their practicality for wearable applications. This research paves the way for accessible, low-power, and high-accuracy AF detection in real-world settings, enabling early diagnosis and timely medical intervention without the need for continuous clinical supervision.
La detección temprana y precisa de la fibrilación auricular (FA) es fundamental para una intervención oportuna y la reducción de riesgos asociados. La tecnología wearable ha surgido como una solución viable para el monitoreo continuo de la FA, pero la implementación de modelos precisos de detección en dispositivos con recursos limitados sigue siendo un desafío debido a las altas demandas computacionales y de memoria. Este estudio propone un enfoque de aprendizaje profundo ligero y eficiente para el diagnostico en tiempo real de la FA en dispositivos portátiles. Diseñamos una serie de modelos de redes neuronales convolucionales (CNN) optimizados para alcanzar alta precisión a la vez que mantenían un bajo costo computacional. Para mejorar aun mas la eficiencia, exploramos técnicas de compresión de aprendizaje profundo, incluyendo poda, cuantización y destilación de conocimiento. Nuestros resultados demuestran que los modelos propuestos logran una precisión de ultima generación mientras reducen significativamente el uso de memoria y la complejidad computacional, lo que los hace adecuados para su implementación en dispositivos de borde. Además, validamos su viabilidad implementándolos en un microcontrolador, demostrando su su aplicabilidad en wearables. Esta investigación abre el camino para una detección de FA accesible, de bajo consumo y de alta precisión en entornos reales, permitiendo un diagnostico temprano y una intervención medica oportuna sin necesidad de supervisión clínica continua.
References
[1] C. A. Morillo et al., “Atrial fibrillation: The current epidemic,” J. Ger. Card., vol. 14, no. 3, p. 195, 2017. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/28592963/
[2] J. Primo et al., “Prevalence of paroxysmal atrial fibrillation in a population assessed by continuous 24-hour monitoring,” Rev. Port. Card., vol. 36, no. 7-8, pp. 535–546, 2017. https://doi.org/10.1016/j.repce.2016.11.013
[3] C. Y. Chenggong Xie, Zhao Wang, J. Liu, and H. Liang, “Machine learning for detecting atrial fibrillation from ecgs: Systematic review and metaanalysis,” RCM, vol. 25, no. 1, p. 8, 2024. https://doi.org/10.31083/j.rcm2501008
[4] A. Y. Hannun et al., “Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network,” Nature Med., vol. 25, no. 1, pp. 65–69, 2019. http://dx.doi.org/10.1038/s41591-018-0268-3
[5] C. A. Fajardo, A. S. Parra, T. Castellanos-Paradaa, and K. Roy, “Low-complexity deep learning models for accurate atrial fibrillation diagnosis,” in 26th Eur. Conf. Art. Intel., 2023. [Online]. Available: https://ecai2023.eu/conf-data/ecai2023/files/STAIRS/stairs2023_05.pdf
[6] L. Biel, O. Pettersson, L. Philipson, and P. Wide, “Ecg analysis: a new approach in human identification,” IEEE Trans. Inst. Measure, vol. 50, no. 3, pp. 808–812, 2001. https://doi.org/10.1109/19.930458
[7] R. Avanzato and F. Beritelli, “Automatic ecg diagnosis using convolutional neural network,” Electronics, vol. 9, no. 6, p. 951, 2020. [Online]. Available: https://www.mdpi.com/2079-9292/9/6/951 DOI: https://doi.org/10.3390/electronics9060951
[8] J. A. Castillo, Y. C. Granados, and C. A. Fajardo, “Patient-specific detection of atrial fibrillation in segments of ecg signals using deep neural networks,” Cien. Ing. Neogranadina, vol. 30, no. 1, pp. 45–58, 2020. https://doi.org/10.18359/rcin.4156
[9] S. Tan, G. Androz, A. Chamseddine, P. Fecteau, A. Courville, Y. Bengio, and J. P. Cohe “Icentia11k: An unsupervised representation learning dataset for arrhythmia subtype discovery,” arXiv preprint arXiv:1910.09570, 2019. [Online]. Available: https://arxiv.org/abs/1910.09570
[10] CardioSTAT, “Ambulatory Cardiac Monitoring,” 2021. [Online]. Available: https://www.cardiostat.com/
[11] P. Rajpurkar, A. Y. Hannun, M. Haghpanahi, C. Bourn, and A. Y. Ng, “Cardiologist-level arrhythmia detection with convolutional neural networks,” arXiv preprint: 1707.01836, 2017. [Online]. Available: https://arxiv.org/abs/1707.01836
[12] Z. Xiong, M. K. Stiles, and J. Zhao, “Robust ECG signal classification for detection of atrial fibrillation using a novel neural network,” Comp. Card., vol. 44, pp. 1–4, 2017. https://doi.org/10.22489/CinC.2017.066-138
[13] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comp. Vis. Pattern Recog., 2016, pp. 770–778. https://doi.org/10.1109/CVPR.2016.90
[14] G. C. Marino´ , A. Petrini, D. Malchiodi, and M. Frasca, “Deep neural networks compression: A comparative survey and choice recommendations,” Neurocomputing, vol. 520, pp. 152–170, 2023. https://doi.org/10.1016/j.neucom.2022.11.072
[15] F. N. Iandola, S. Han, M. W. Moskewicz, K. Ashraf, W. J. Dally, and K. Keutzer, “SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 mb model size,” arXiv preprint: 1602.07360, 2016. [Online]. Available: https://arxiv.org/abs/1602.07360
[16] C. Bucilua, R. Caruana, and A. Niculescu- Mizil, “Model compression,” in Proc. 12th ACM SIGKDD Int. Conf. Know. Disc. Data Mining, 2006, pp. 535–541. https://doi.org/10.1145/1150402.1150464
[17] G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint: 1503.02531, 2015. [Online]. Available: http://arxiv.org/abs/1503.02531
[18] S. Han, J. Pool, J. Tran, and W. J. Dally, “Learning both weights and connections for efficient neural networks,” arXiv preprint: 1506.02626, pp. 1–9, 2015. [Online]. Available: http://arxiv.org/abs/1506.02626
[19] B. Jacob, S. Kligys, B. Chen, M. Zhu, M. Tang, A. Howard, H. Adam, and D. Kalenichenko, “Quantization and training of neural networks for efficient integerarithmetic- only inference,” in Proc. IEEE Conf. Comp. Vis. Pattern Recog., 2018, pp. 2704–2713. https://doi.org/10.1109/CVPR.2018.00286
[20] SparkFun-Electronics, “Teensy 4.1.” [Online]. Available: https://www.sparkfun.com/products/16771
[21] O. A. Ademola, M. Leier, and E. Petlenkov, “Evaluation of deep neural network compression methods for edge devices using weighted scorebased ranking scheme,” Sensors, vol. 21, no. 22, p. 7529, 2021. https://doi.org/10.3390/s21227529
[22] J. Rubin, S. Parvaneh, A. Rahman, B. Conroy, and S. Babaeizadeh, “Densely connected convolutional networks for detection of atrial fibrillation from short single-lead ecg recordings,” J. Electrocard., vol. 51, no. 6, pp. S18–S21, 2018. https://doi.org/10.1016/j.jelectrocard.2018.08.008
[23] Q. Yao, R. Wang, X. Fan, J. Liu, and Y. Li, “Multi-class arrhythmia detection from 12-lead variedlength ECG using attention-based time-incremental convolutional neural network,” Info. Fus., vol. 53, no. June 2019, pp. 174–182, 2020. https://doi.org/10.1016/j.inffus.2019.06.024
[24] K. Fonseca, S. Osorio, J. Castillo, and C. Fajardo, “Contrastive learning for atrial fibrillation detection in challenging scenarios,” in 2022 30th Eur. Signal Proc. Conf. (EUSIPCO). IEEE, 2022, pp. 1218–1222. https://doi.org/10.23919/EUSIPCO55093.2022.9909842
[25] A. R. Iskandar, S. Jundi, H. Rifaı, and N. Rizoug, “Ecg classification using an optimal temporal convolutional network for remote health monitoring,” Sensors, vol. 23, no. 3, p. 1697, 2023. https://doi.org/10.3390/s23031697
[26] S. Asgari et al., “Automatic detection of atrial fibrillation using stationary wavelet transform and support vector machine,” Comp. Biol. Med., vol. 60, pp. 132–142, 2015. https://doi.org/10.1016/j.compbiomed.2015.03.005
[27] R. J. Martis et al., “Automated detection of atrial fibrillation using bayesian paradigm,” Konwledgebased Syst., vol. 54, pp. 269–275, 2013. https://doi.org/10.1016/j.knosys.2013.09.016
[28] M. Mohebbi and H. Ghassemian, “Detection of atrial fibrillation episodes using svm,” in 2008 30th Ann. Int. Conf. IEEE Eng. Med. Biol. Soc. IEEE, 2008, pp. 177–180. https://doi.org/10.1109/IEMBS.2008.4649119
[29] S. Parvaneh, J. Rubin, S. Babaeizadeh, and M. Xu-Wilson, “Cardiac arrhythmia detection using deep learning: A review,” J. Electrocard., vol. 57, pp. S70–S74, 2019. https://doi.org/10.1016/j.jelectrocard.2019.08.004
[30] A. H. Ribeiro, D. M. Tavares de Oliveira, P. R. Gomes, D. Canazart, M. Ferreira, C. R. Andersson, A. L. Ribeiro, T. B. Schon, W. Meira, and L. P. Rocha, “Automatic diagnosis of the 12-lead ecg using a deep neural network,” Nature Comm., vol. 11, no. 1, p. 1760, 2020. https://doi.org/10.1038/s41467-020-15432-4
[31] O. Yildirim, U. B. Baloglu, R. Tan, and U. R. Acharya, “Arrhythmia classification using transformer-based deep learning model,” Biomed. Signal Proc. Control, vol. 81, p. 104477, 2023.
[32] J. Rubin, S. Parvaneh, A. Rahman, B. Conroy, and S. Babaeizadeh, “Densely connected convolutional networks for ecg classification,” J. Electrocard., vol. 57, pp. S20–S24, 2018.
[33] Y. Oh, H. Moon, D. Kim, J. Park, and D. Kim, “Transformer-based arrhythmia detection on ecg signals,” Sensors, vol. 22, no. 8, p. 2966, 2022.
[34] I. J. Selvam, M. Madhavan, and S. K. Kumarasamy, “Detection and classification of electrocardiography using hybrid deep learning models,” Hellenic J. Card., vol. 81, pp. 75–84, 2025. https://doi.org/10.1016/j.hjc.2024.08.011
[35] A. R. Ismail, S. Jovanovic, N. Ramzan, and H. Rabah, “Ecg classification using an optimal temporal convolutional network for remote health monitoring,” Sensors, vol. 23, no. 3, p. 1697, 2023. https://doi.org/10.3390/s23031697
[36] T. Mahmud, A. Barua, D. Islam, M. S. Hossain, R. Chakma, K. Barua, M. Monju, and K. Andersson, “Ensemble deep learning approach for ECG-based cardiac disease detection: Signal and image analysis,” in 2023 Int. Conf. Info. Comm. Tech. Sust. Dev. (ICICT4SD), 2023, pp. 70–74. https://doi.org/10.1109/ICICT4SD59951.2023.10303625
[37] Z. Zhao, X. Dong, J. Liu, Y. Guo, and Y. Yao, “ECG classification using a lightweight transformer-based model,” Biomed Signal Proc. Control, vol. 78, p. 103931, 2022.
[38] E. Tanghatari, M. Kamal, A. Afzali-Kusha, and M. Pedram, “Federated learning by employing knowledge distillation on edge devices with limited hardware resources, Neurocomputing, vol. 531, pp. 87–99, 2023. https://doi.org/10.1016/j.neucom.2023.02.011
How to Cite
APA
ACM
ACS
ABNT
Chicago
Harvard
IEEE
MLA
Turabian
Vancouver
Download Citation
CrossRef Cited-by
1. Yash Akbari, Ningrong Lei, Nilesh Patel, Yonghong Peng, Oliver Faust. (2025). Atrial Fibrillation Detection on the Embedded Edge: Energy-Efficient Inference on a Low-Power Microcontroller. Sensors, 25(21), p.6601. https://doi.org/10.3390/s25216601.
Dimensions
PlumX
Article abstract page views
Downloads
License
Copyright (c) 2025 Carlos A. Fajardo, Andrés S. Parra, Tania V. Castellanos-Parada
This work is licensed under a Creative Commons Attribution 4.0 International License.
The authors or holders of the copyright for each article hereby confer exclusive, limited and free authorization on the Universidad Nacional de Colombia's journal Ingeniería e Investigación concerning the aforementioned article which, once it has been evaluated and approved, will be submitted for publication, in line with the following items:
1. The version which has been corrected according to the evaluators' suggestions will be remitted and it will be made clear whether the aforementioned article is an unedited document regarding which the rights to be authorized are held and total responsibility will be assumed by the authors for the content of the work being submitted to Ingeniería e Investigación, the Universidad Nacional de Colombia and third-parties;
2. The authorization conferred on the journal will come into force from the date on which it is included in the respective volume and issue of Ingeniería e Investigación in the Open Journal Systems and on the journal's main page (https://revistas.unal.edu.co/index.php/ingeinv), as well as in different databases and indices in which the publication is indexed;
3. The authors authorize the Universidad Nacional de Colombia's journal Ingeniería e Investigación to publish the document in whatever required format (printed, digital, electronic or whatsoever known or yet to be discovered form) and authorize Ingeniería e Investigación to include the work in any indices and/or search engines deemed necessary for promoting its diffusion;
4. The authors accept that such authorization is given free of charge and they, therefore, waive any right to receive remuneration from the publication, distribution, public communication and any use whatsoever referred to in the terms of this authorization.
