Transformer-Driven Cardiac Image Segmentation: A Swin-EchoNet Approach for Two-Chamber Echocardiography
DOI:
https://doi.org/10.3991/ijoe.v22i07.60655Keywords:
Swin Transformer, Deep Learning, Echocardiography, Cardiac Chamber Segmentation, Two-Chamber View, Vision Transformer, Attention Mechanism, Medical Image SegmentationAbstract
Echocardiographic segmentation involves the automated delineation of cardiac chambers from ultrasound images to enable quantitative assessment of cardiac structure and function, which is essential for diagnosing and managing cardiovascular diseases. However, existing CNN-based segmentation models often struggle to preserve fine boundary details and global spatial context due to speckle noise, low contrast, and imaging artifacts inherent to echocardiography. To overcome these limitations, this study proposes Swin-EchoNet, a transformer-based segmentation framework that combines shifted-window self-attention with hierarchical feature extraction to capture both local texture information and long-range dependencies. The proposed approach was trained and evaluated on two distinct datasets: the CAMUS dataset comprising 1,800 echocardiographic images and a clinical Aster-MIMS dataset consisting of 1,550 frames collected from hospital studies. On the CAMUS dataset, Swin-EchoNet achieved a mean Dice score of 0.951 ± 0.227, while on the Aster-MIMS dataset, it obtained an average Dice coefficient of 0.9738 with a mean Intersection over Union (IoU) of 0.7263. Comparative evaluations demonstrate that Swin-EchoNet outperforms U-Net, ResUNet, and Attention U-Net, highlighting its robustness and clinical applicability.
References
[1] A. Naveed, R. Atique, H. A. Saeed, J. Sharif, A. Haidar, and A. Samad, “Cardiovascular diseases: Understanding the leading cause of death worldwide,” Glob. J. Multidiscip. Sci. Arts, vol. 1, no. 1, pp. 100–110, 2024.
[2] V. Christoffels and B. Jensen, “Cardiac morphogenesis: Specification of the four-chambered heart,” Cold Spring Harb. Perspect. Biol., vol. 12, no. 10, p. a037143, 2020. https://doi.org/10.1101/cshperspect.a037143
[3] M. F. Buijtendijk, P. Barnett, and M. J. van den Hoff, “Development of the human heart,” Am. J. Med. Genet. C Semin. Med. Genet., vol. 184, no. 1, pp. 7–22, 2020. https://doi.org/10.1002/ajmg.c.31778
[4] C. Liguori et al., “Role of CT and MRI in cardiac emergencies,” Tomography, vol. 8, no. 3, pp. 1386–1400, 2022. https://doi.org/10.3390/tomography8030112
[5] L. D. Weberling, D. Lossnitzer, N. Frey, and F. Andre, “Coronary computed tomography vs. cardiac magnetic resonance imaging in the evaluation of coronary artery disease,” Diagnostics, vol. 13, no. 1, p. 125, 2022.https://doi.org/10.3390/diagnostics13010125
[6] M. Habijan et al., “Overview of the whole heart and heart chamber segmentation methods,” Cardiovasc. Eng. Technol., vol. 11, no. 6, pp. 725–747, 2020.
[7] L. Lenchik et al., “Automated segmentation of tissues using CT and MRI: A systematic review,” Acad. Radiol., vol. 26, no. 12, pp. 1695–1706, 2019. https://doi.org/10.1016/j.acra.2019.07.006
[8] A. C. Kwan, G. Salto, S. Cheng, and D. Ouyang, “Artificial intelligence in computer vision: Cardiac MRI and multimodality imaging segmentation,” Curr. Cardiovasc. Risk Rep., vol. 15, no. 9, p. 18, 2021. https://doi.org/10.1007/s12170-021-00678-4
[9] A. Amer, X. Ye, and F. Janan, “ResDUnet: A deep learning-based left ventricle segmentation method for echocardiography,” IEEE Access, vol. 9, pp. 159755–159763, 2021. https://doi.org/10.1109/ACCESS.2021.3122256
[10] N. Azarmehr et al., “Segmentation of left ventricle in 2D echocardiography using deep learning,” in Proc. MIUA, pp. 497–504, 2019. https://doi.org/10.3389/fcvm.2022.937068
[11] C. Chen et al., “Deep learning for cardiac image segmentation: A review,” Front. Cardiovasc. Med., vol. 7, p. 25, 2020. https://doi.org/10.3389/fcvm.2020.00025
[12] L. Wu et al., “Automatic segmentation of the left ventricle in apical four-chamber view on transesophageal echocardiography based on UNeXt,” Diagnostics, vol. 14, no. 23, p. 2766, 2024. https://doi.org/10.3390/diagnostics14232766
[13] H. Assadi et al., “AI-derived segmentation of four-chamber cine cardiac magnetic resonance,” Eur. Radiol. Exp., vol. 8, no. 1, p. 77, 2024. https://doi.org/10.1186/s41747-024-00477-7
[14] N. Fatima, S. Afrakhteh, G. Iacca, and L. Demi, “Automatic segmentation of 2-D echocardiography ultrasound images using GAN,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 71, no. 11, pp. 1552–1564, 2024. https://doi.org/10.1109/TUFFC.2024.3393026
[15] J. Wang, X. Bo, G. Li, and Y. Tan, “Ventricular segmentation algorithm for echocardiography based on transfer learning and GAN,” Syst. Soft Comput., vol. 6, p. 200139, 2024. https://doi.org/10.1016/j.sasc.2024.200139
[16] Z. Ye et al., “HFE-mamba: High-frequency enhanced mamba network for left ventricle segmentation,” IEEE Access, 2024. https://doi.org/10.1109/ACCESS.2024.3424546
[17] C. J. Chao et al., “Foundation versus domain-specific models for left ventricular segmentation,” npj Digit. Med., vol. 8, no. 1, p. 341, 2025. https://doi.org/10.1038/s41746-025-01730-y
[18] S. Kang et al., “Left ventricle segmentation in transesophageal echocardiography,” PLoS One, vol. 18, no. 1, p. e0280485, 2023. https://doi.org/10.1371/journal.pone.0280485
[19] M. J. Mortada et al., “Segmentation of left heart anatomical structures using deep learning,” Diagnostics, vol. 13, no. 10, p. 1683, 2023. https://doi.org/10.3390/diagnostics13101683
[20] H. Li et al., “EchoEFNet: Multi-task deep learning network for automatic calculation of left ventricular ejection fraction,” Comput. Biol. Med., vol. 156, p. 106705, 2023. https://doi.org/10.1016/j.compbiomed.2023.106705
[21] G. F. Cacao, D. Du, and N. Nair, “Unsupervised image segmentation on 2D echocardiogram,” Algorithms, vol. 17, no. 11, p. 515, 2024. https://doi.org/10.3390/a17110515
[22] J. Ding and X. Han, “Semi-supervised echocardiography segmentation using image and frequency enhancement,” IEEE Access, vol. 12, pp. 92549–92559, 2024. https://doi.org/10.1109/ACCESS.2024.3408952
[23] P. Naidoo et al., “Consensus-guided evaluation of self-supervised learning in echocardiographic segmentation,” Comput. Biol. Med., vol. 198, p. 111148, 2025. https://doi.org/10.1016/j.compbiomed.2025.111148
[24] A. Chernyshov et al., “Automated segmentation and quantification of the right ventricle in 2-D echocardiography,” Ultrasound Med. Biol., vol. 50, no. 4, pp. 540–548, 2024. https://doi.org/10.1016/j.ultrasmedbio.2023.12.018
[25] Z. Qian, T. Hu, J. Wang, and Z. Yang, “U-shape-based network for left ventricular segmentation with contrastive pretraining,” Sci. Rep., vol. 14, no. 1, p. 29689, 2024. https://doi.org/10.1038/s41598-024-81523-7
[26] S. Nemri and L. Duong, “Shifted windows vision transformer for echocardiographic image segmentation,” Biomed. Phys. Eng. Express, vol. 10, no. 6, p. 065017, 2024. https://doi.org/10.1088/2057-1976/ad7594
[27] S. Hasan, “CAMUS human heart dataset,” Kaggle, 2024.
[28] S. Leclerc et al., “Deep learning for segmentation using an open large-scale dataset in 2D echocardiography,” IEEE Trans. Med. Imaging, vol. 38, no. 9, pp. 2198–2210, 2019. https://doi.org/10.1109/TMI.2019.2900516
[29] R. M. Lang et al., “Recommendations for cardiac chamber quantification by echocardiography,” Eur. Heart J. Cardiovasc. Imaging, vol. 16, no. 3, pp. 233–271, 2015. https://doi.org/10.1093/ehjci/jev014
[30] Z. Liu et al., “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proc. IEEE/CVF ICCV, pp. 10012–10022, 2021.
[31] V. Thamilarasi and R. Roselin, “U-Net for lung image segmentation and classification,” INFOCOMP J. Comput. Sci., vol. 20, no. 1, pp. 101–108, 2021.
[32] G. Gaál, B. Maga, and A. Lukács, “Attention U-Net based adversarial architectures for chest X-ray lung segmentation,” arXiv, p. arXiv:2003.10304, 2020. https://doi.org/10.48550/arXiv.2003.10304
[33] F. I. Diakogiannis, F. Waldner, P. Caccetta, and C. Wu, “ResUNet-a: A deep learning framework for semantic segmentation,” ISPRS J. Photogramm. Remote Sens., vol. 162, pp. 94–114, 2020 https://doi.org/10.1016/j.isprsjprs.2020.01.013
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Shamla Beevi A., Amit Nagarkoti, Saidalavi Kalady
This work is licensed under a Creative Commons Attribution 4.0 International License.
