Semi-Supervised Left Atrium Segmentation in 3D Cardiac MRI Using Confidence-Guided Pseudo-Labeling
Article Sidebar
Main Article Content
Abstract
This study proposes a semi-supervised learning framework for left atrium (LA) segmentation from three-dimensional cardiac Magnetic Resonance Imaging (MRI) using pseudo-labeling. The objective is to improve segmentation performance under limited labeled-data conditions. The proposed method integrates a 3D U-Net architecture with an iterative training pipeline and dynamic confidence-based pseudo-label refinement. Using the Medical Segmentation Decathlon dataset, experiments demonstrate that the semi-supervised model achieves a mean Dice Coefficient (DSC) of 0.9066 ± 0.0043 and a mean Average Hausdorff Distance (AHD) of 2.2409 ± 0.3661, surpassing the fully supervised baseline (DSC: 0.8519 ± 0.0395; AHD: 4.7696 ± 1.3128). Qualitative evaluation further confirms reduced false positives and enhanced anatomical precision. The results indicate that the proposed approach effectively leverages unlabeled data to achieve high segmentation accuracy with minimal manual annotation, providing a practical solution for clinical management of atrial fibrillation (AF).
Article Details
References
Antonelli, M., Reinke, A., Bakas, S., Farahani, K., Kopp-Schneider, A., Landman, B. A., et al. (2022). The medical segmentation decathlon. Nature communications, 13(1), 4128.
Aryan, R., Kejriwal, V., Patel, V., Aggarwal, A., Khanna, V., Thomas, S. B., et al. (2022, November). Left Atrium Segmentation Using Deep Learning Model. The Proceedings of the 19th 2022 IEEE India Council International Conference (INDICON), 1-5. IEEE.
Bardis, M., Houshyar, R., Chantaduly, C., Ushinsky, A., Glavis-Bloom, J., Shaver, M., et al. (2020). Deep learning with limited data: organ segmentation performance by U-Net. Electronics, 9(8), 1199.
Kiryati, N., and Landau, Y. (2021). Dataset growth in medical image analysis research. Journal of imaging, 7(8), 155.
Krittayaphong, R., Rangsin, R., Thinkhamrop, B., Hurst, C., Rattanamongkolgul, S., Sripaiboonkij, N., et al. (2016). Prevalence and associating factors of atrial fibrillation in patients with hypertension: a nation-wide study. BMC Cardiovascular Disorders, 16(1), 57.
Liu, Y., Wang, W., Luo, G., Wang, K., and Li, S. (2022). A contrastive consistency semi-supervised left atrium segmentation model. Computerized Medical Imaging and Graphics, 99, 102092.
Müller, D., Soto-Rey, I., and Kramer, F. (2022). Towards a guideline for evaluation metrics in medical image segmentation. BMC Research Notes, 15(1), 210.
Oltman, C. G., Kim, T. P., Lee, J. W., Lupu, J. D., Zhu, R., and Moussa, I. D. (2024). Prevalence, Management, and Comorbidities of Adults With Atrial Fibrillation in the United States, 2019 to 2023. JACC: Advances, 3(11), 101330.
Ottesen, J. A., Tong, E., Emblem, K. E., Latysheva, A., Zaharchuk, G., Bjørnerud, A., et al. (2025). Semi-Supervised Learning Allows for Improved Segmentation with Reduced Annotations of Brain Metastases Using Multicenter MRI Data. Journal of Magnetic Resonance Imaging, 61(6), 2469-2479.
Qu, C., Zhang, T., Qiao, H., Tang, Y., Yuille, A. L., and Zhou, Z. (2023). Abdomenatlas-8k: Annotating 8,000 ct volumes for multi-organ segmentation in three weeks. Advances in Neural Information Processing Systems, 36, 36620-36636.
Roth, G. A., Mensah, G. A., Johnson, C. O., Addolorato, G., Ammirati, E., Baddour, L. M., et al. (2020). Global burden of cardiovascular diseases and risk factors, 1990–2019: update from the GBD 2019 study. Journal of the American college of cardiology, 76(25), 2982-3021.
Shi, Z., Jiang, M., Li, Y., Wei, B., Wang, Z., Wu, Y., et al. (2024). MLC: Multi-level consistency learning for semi-supervised left atrium segmentation. Expert Systems with Applications, 244, 122903.
Swetha, S., Rafee, A., Manjula, S. H., and Venugopal, K. R. (2023, December). Optimizing Left Atrium Segmentation: A Modified U-NET Architecture with MRI Image Slicing. In 2023 IEEE 2nd International Conference on Data, Decision and Systems (ICDDS) (pp. 1-6). IEEE.
Suwanwela, N. C., Chutinet, A., Autjimanon, H., Ounahachok, T., Decha-Umphai, C., Chockchai, S., et al. (2021). Atrial fibrillation prevalence and risk profile from novel community-based screening in Thailand: A prospective multi-centre study. IJC Heart and Vasculature, 32, 100709.
Taha, A. A., and Hanbury, A. (2015). Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC medical imaging, 15(1), 29.
Uslu, F., Varela, M., Boniface, G., Mahenthran, T., Chubb, H., and Bharath, A. A. (2021). LA-Net: A multi-task deep network for the segmentation of the left atrium. IEEE transactions on medical imaging, 41(2), 456-464.
Wang, J., Liu, X., Yin, J., and Ding, P. (2022). DC-net: Dual-Consistency semi-supervised learning for 3D left atrium segmentation from MRI. Biomedical Signal Processing and Control, 78, 103870.
Xu, G., Qian, X., Shao, H. C., Luo, J., Lu, W., and Zhang, Y. (2024). A SAM-guided and Match-based Semi-Supervised Segmentation Framework for Medical Imaging. arXiv preprint arXiv:2411.16949.
Yang, Q., Wang, C., Pan, K., Xia, B., Xie, R., and Shi, J. (2024). An improved 3D-UNet-based brain hippocampus segmentation model based on MR images. BMC Medical Imaging, 24(1), 166.
Zhang, X., Noga, M., Martin, D. G., and Punithakumar, K. (2021). Fully automated left atrium segmentation from anatomical cine long-axis MRI sequences using deep convolutional neural network with unscented Kalman filter. Medical image analysis, 68, 101916.
