Fetal heart abnormalities, particularly fetal heart arrhythmias, are becoming an increasing concern in maternal healthcare. Early and accurate detection remains challenging due to limitations in conventional diagnostic techniques and inadequate signal processing methods. Existing studies have applied various machine learning and deep learning approaches; however, capturing the complex characteristics of fetal electrocardiogram (ECG) signals remains a significant gap. This study proposes an advanced deep learning-based approach for detecting fetal heart arrhythmias using non-invasive ECG signals. Initially, a preprocessing step is performed using blind source separation to isolate fetal ECG signals from maternal ECG within a frequency range of 5 Hz to 15 Hz. Subsequently, segmentation is applied to extract fixed-length ECG signal segments for analysis. The proposed model integrates a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) architecture. The CNN component extracts spatial and morphological features from ECG signals, while the LSTM network captures temporal dependencies and time-frequency relationships across the signal. The model was evaluated using the NIFEADB (Non-Invasive Fetal ECG Arrhythmia Database) obtained from PhysioNet. The proposed method achieved a high classification accuracy of 99.2%, demonstrating its effectiveness in identifying fetal cardiac abnormalities. This approach provides a promising, non-invasive, and reliable tool for early detection of fetal arrhythmias, potentially improving prenatal diagnosis and clinical decision-making.

Keywords:Fetal ECG, heart arrhythmia, deep learning, CNN-LSTM, signal processing

🔒 This is a subscription article

Full text is available to subscribers and institutional members. Please choose an option below to access it.

1. World Health Organization. Prevention and surveillance of birth defects: Report of a meeting of regional programme managers; 14–16 April 2015; New Delhi, India. Available from: https://iris.who.int/server/api/core/bitstreams/b8ed331e-1673-426c-9a01-b038c6a026da/content
2. Martin SS, Aday AW, Almarzooq ZI, Anderson CAM, Arora P, Avery CL, et al. 2024 heart disease and stroke statistics: a report of US and global data from the American Heart Association. Circulation. 2024 Feb 20;149(8):e347–e913. https://doi.org/10.1161/cir.0000000000001209.
3. Strasburger JF, Cuneo BF, Michon MM, Gotteiner NL, Deal BJ, McGregor SN, et al. Amiodarone therapy for drug-refractory fetal tachycardia. Circulation. 2004 Jan 27;109(3):375–9. doi: 1161 /01.CIR.0000109494.05317.58. PMID: 14732753.
4. Puertas A, Góngora J, Valverde M, Revelles L, Manzanares S, Carrillo MP. Cardiotocography alone vs cardiotocography with ST segment analysis for intrapartum fetal monitoring in women with late-term pregnancy: A randomized controlled trial. Eur J Obstet Gynecol Reprod Biol. 2019 Mar;234:213–217. doi: 1016/j.ejogrb.2019.01.023. PMID: 30731334.
5. Quadri R, Khanam S. Analysing and evaluating the performance of deep-learning-based arrhythmia detection using electrocardiogram signals. Int J Adv Res Comput Sci. 2024;15(2):74. Available from: https://ijarcs.info/index.php/Ijarcs/article/view/7065
6. Wang, L., Zhao, C., Dong, M., & Ota, K. (2022). Fetal ECG Signal Extraction From Long-Term Abdominal Recordings Based on Adaptive QRS Removal and Joint Blind Source Separation. IEEE Sensors Journal, 22, 20718-20729.
7. Liu J, Xu H, Wang J, Peng X, He C. Non-invasive diagnosis of fetal arrhythmia based on multi-domain feature and hierarchical extreme learning machine. Biomed Signal Process Control. 2023 Jan;79(Pt 2):104191.
8. Kwon D, Kang H, Lee D, Kim YC. Deep learning-based prediction of atrial fibrillation from polar transformed time-frequency electrocardiogram. PLoS One. 2025 Mar 10;20(3):e0317630. doi: 1371/journal.pone.0317630. PMID: 40063554; PMCID: PMC11892834.
9. Asgarabad MohammadReza Einollahi. A comprehensive comparative analysis of machine learning algorithms in heart disease prediction. Discov Artif Intell. 2026. doi: 1007/s44163-026-01048-y
10. Zhang Y, GuA, Xiao Z, Xing Y, Yang C, Li J, et al., et al. Wearable fetal ECG monitoring system from abdominal electrocardiography recording. Biosensors (Basel). 2022 Jun 30;12(7):475. doi: 3390/bios12070475. PMID: 35884277; PMCID: PMC9313261.
11. Veenadevi SV, Padmavathi C, Shanthamma B, Abbigeri BG, Pavithra KM. Extraction of fetal electrocardiogram from maternal electrocardiogram and classification of normal and abnormal signals. In: 2017 IEEE 2nd International Conference on Signal and Image Processing (ICSIP); 2017. p. 396–401.
12. Niknazar M, Becker H, Rivet B, Jutten C, Comon P. Blind source separation of underdetermined mixtures of event-related sources. Signal Process. 2014;101:52–64. doi: 1016/j.sigpro.2014.01.031.
13. Prasad, L., & Iyengar, S.S. (1997). Wavelet Analysis with Applications to Image Processing (1st ed.). CRC Press. https://doi.org/10.1201/9780367811310.
14. Nakatani S, Yamamoto K, Ohtsuki T. Fetal arrhythmia detection based on labeling considering heartbeat interval. Bioengineering (Basel). 2022 Dec 30;10(1):48. doi: 3390/bioengineering 10010048. PMID: 36671621; PMCID: PMC9855115.
15. Rai RK, Singh A, Srivastva R, Kumar G. Fetal ECG arrhythmia detection based on DensNet transfer learning. Front Biomed Technol. 2023;10(4):417–426.
16. Al-Selwi SM, Hassan MF, Abdulkadir SJ, et al. RNN-LSTM: From applications to modeling techniques and beyond—Systematic review. J King Saud Univ Comput Inf Sci. Volume 36, Issue 5, June 2024, 102068.
17. Dash SS, Nath MK. Non-invasive techniques for fECG analysis in fetal heart monitoring: A systematic review. Signals. 2025;6(4):61. doi: 3390/signals6040061.

Fetal heart abnormalities, particularly fetal heart arrhythmias, are becoming an increasing concern in maternal healthcare. Early and accurate detection remains challenging due to limitations in conventional diagnostic techniques and inadequate signal processing methods. Existing studies have applied various machine learning and deep learning approaches; however, capturing the complex characteristics of fetal electrocardiogram (ECG) signals remains a significant gap. This study proposes an advanced deep learning-based approach for detecting fetal heart arrhythmias using non-invasive ECG signals. Initially, a preprocessing step is performed using blind source separation to isolate fetal ECG signals from maternal ECG within a frequency range of 5 Hz to 15 Hz. Subsequently, segmentation is applied to extract fixed-length ECG signal segments for analysis. The proposed model integrates a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) architecture. The CNN component extracts spatial and morphological features from ECG signals, while the LSTM network captures temporal dependencies and time-frequency relationships across the signal. The model was evaluated using the NIFEADB (Non-Invasive Fetal ECG Arrhythmia Database) obtained from PhysioNet. The proposed method achieved a high classification accuracy of 99.2%, demonstrating its effectiveness in identifying fetal cardiac abnormalities. This approach provides a promising, non-invasive, and reliable tool for early detection of fetal arrhythmias, potentially improving prenatal diagnosis and clinical decision-making.

Keywords:Fetal ECG, heart arrhythmia, deep learning, CNN-LSTM, signal processing

🔒 This is a subscription article

Full text is available to subscribers and institutional members. Please choose an option below to access it.

1. World Health Organization. Prevention and surveillance of birth defects: Report of a meeting of regional programme managers; 14–16 April 2015; New Delhi, India. Available from: https://iris.who.int/server/api/core/bitstreams/b8ed331e-1673-426c-9a01-b038c6a026da/content
2. Martin SS, Aday AW, Almarzooq ZI, Anderson CAM, Arora P, Avery CL, et al. 2024 heart disease and stroke statistics: a report of US and global data from the American Heart Association. Circulation. 2024 Feb 20;149(8):e347–e913. https://doi.org/10.1161/cir.0000000000001209.
3. Strasburger JF, Cuneo BF, Michon MM, Gotteiner NL, Deal BJ, McGregor SN, et al. Amiodarone therapy for drug-refractory fetal tachycardia. Circulation. 2004 Jan 27;109(3):375–9. doi: 1161 /01.CIR.0000109494.05317.58. PMID: 14732753.
4. Puertas A, Góngora J, Valverde M, Revelles L, Manzanares S, Carrillo MP. Cardiotocography alone vs cardiotocography with ST segment analysis for intrapartum fetal monitoring in women with late-term pregnancy: A randomized controlled trial. Eur J Obstet Gynecol Reprod Biol. 2019 Mar;234:213–217. doi: 1016/j.ejogrb.2019.01.023. PMID: 30731334.
5. Quadri R, Khanam S. Analysing and evaluating the performance of deep-learning-based arrhythmia detection using electrocardiogram signals. Int J Adv Res Comput Sci. 2024;15(2):74. Available from: https://ijarcs.info/index.php/Ijarcs/article/view/7065
6. Wang, L., Zhao, C., Dong, M., & Ota, K. (2022). Fetal ECG Signal Extraction From Long-Term Abdominal Recordings Based on Adaptive QRS Removal and Joint Blind Source Separation. IEEE Sensors Journal, 22, 20718-20729.
7. Liu J, Xu H, Wang J, Peng X, He C. Non-invasive diagnosis of fetal arrhythmia based on multi-domain feature and hierarchical extreme learning machine. Biomed Signal Process Control. 2023 Jan;79(Pt 2):104191.
8. Kwon D, Kang H, Lee D, Kim YC. Deep learning-based prediction of atrial fibrillation from polar transformed time-frequency electrocardiogram. PLoS One. 2025 Mar 10;20(3):e0317630. doi: 1371/journal.pone.0317630. PMID: 40063554; PMCID: PMC11892834.
9. Asgarabad MohammadReza Einollahi. A comprehensive comparative analysis of machine learning algorithms in heart disease prediction. Discov Artif Intell. 2026. doi: 1007/s44163-026-01048-y
10. Zhang Y, GuA, Xiao Z, Xing Y, Yang C, Li J, et al., et al. Wearable fetal ECG monitoring system from abdominal electrocardiography recording. Biosensors (Basel). 2022 Jun 30;12(7):475. doi: 3390/bios12070475. PMID: 35884277; PMCID: PMC9313261.
11. Veenadevi SV, Padmavathi C, Shanthamma B, Abbigeri BG, Pavithra KM. Extraction of fetal electrocardiogram from maternal electrocardiogram and classification of normal and abnormal signals. In: 2017 IEEE 2nd International Conference on Signal and Image Processing (ICSIP); 2017. p. 396–401.
12. Niknazar M, Becker H, Rivet B, Jutten C, Comon P. Blind source separation of underdetermined mixtures of event-related sources. Signal Process. 2014;101:52–64. doi: 1016/j.sigpro.2014.01.031.
13. Prasad, L., & Iyengar, S.S. (1997). Wavelet Analysis with Applications to Image Processing (1st ed.). CRC Press. https://doi.org/10.1201/9780367811310.
14. Nakatani S, Yamamoto K, Ohtsuki T. Fetal arrhythmia detection based on labeling considering heartbeat interval. Bioengineering (Basel). 2022 Dec 30;10(1):48. doi: 3390/bioengineering 10010048. PMID: 36671621; PMCID: PMC9855115.
15. Rai RK, Singh A, Srivastva R, Kumar G. Fetal ECG arrhythmia detection based on DensNet transfer learning. Front Biomed Technol. 2023;10(4):417–426.
16. Al-Selwi SM, Hassan MF, Abdulkadir SJ, et al. RNN-LSTM: From applications to modeling techniques and beyond—Systematic review. J King Saud Univ Comput Inf Sci. Volume 36, Issue 5, June 2024, 102068.
17. Dash SS, Nath MK. Non-invasive techniques for fECG analysis in fetal heart monitoring: A systematic review. Signals. 2025;6(4):61. doi: 3390/signals6040061.