ABSTRACT
Cardiac image segmentation is a sub task of medical image segmentation, which has very important clinical significance in the prevention and diagnosis of cardiac diseases. Recently, segmentation models based on the Transformer architecture have achieved remarkable performance due to their advantages in handling long-range dependencies and capturing global contextual information. However, the Transformer based segmentation model only uses simple unprocessed skip connections to compensate for information loss during the down-sampling process, resulting in its potential in cardiac image segmentation remains largely unexplored. In this work, we propose CardiacSegFormer, which consists of various improvements, such as the application of self-attention modules in skip connections to facilitate better information fusion. At the same time, low-entropy constraint is introduced, cosine annealing warm restart algorithms are used to help training a more effective model. Experiment on Automated Cardiac Diagnosis Challenge (ACDC) datasets demonstrate that the proposed method has a Dice Similarity Coefficient (DSC) of 91.84%, which is superior to the current the-state-of-art model. To conclusion, this work provides a high-performing baseline model for cardiac image segmentation tasks, showcasing its value in the improvement and enhancement of related models.
- Ramesh, K., Kumar, G. K., Swapna, K., Datta, D., Rajest, S. S. J. E. E. T. o. P. H. and Technology A review of medical image segmentation algorithms, 7, 27 (2021), e6-e6.Google Scholar
- Chen, C., Qin, C., Qiu, H., Tarroni, G., Duan, J., Bai, W. and Rueckert, D. J. F. i. C. M. Deep learning for cardiac image segmentation: a review, 7 (2020), 25.Google Scholar
- Long, J., Shelhamer, E. and Darrell, T. Fully convolutional networks for semantic segmentation. City, 2015.Google Scholar
- Ronneberger, O., Fischer, P. and Brox, T. U-net: Convolutional networks for biomedical image segmentation. Springer, City, 2015.Google Scholar
- Yu, L., Cheng, J.-Z., Dou, Q., Yang, X., Chen, H., Qin, J. and Heng, P.-A. Automatic 3D cardiovascular MR segmentation with densely-connected volumetric convnets. Springer, City, 2017.Google ScholarDigital Library
- Li, X., Chen, H., Qi, X., Dou, Q., Fu, C.-W. and Heng, P.-A. J. I. t. o. m. i. H-DenseUNet: hybrid densely connected UNet for liver and tumor segmentation from CT volumes, 37, 12 (2018), 2663-2674.Google Scholar
- Yu, Q., Xie, L., Wang, Y., Zhou, Y., Fishman, E. K. and Yuille, A. L. Recurrent saliency transformation network: Incorporating multi-stage visual cues for small organ segmentation. City, 2018.Google Scholar
- Zhou, Y., Xie, L., Shen, W., Wang, Y., Fishman, E. K. and Yuille, A. L. A fixed-point model for pancreas segmentation in abdominal CT scans. Springer, City, 2017.Google ScholarDigital Library
- Çiçek, Ö., Abdulkadir, A., Lienkamp, S. S., Brox, T. and Ronneberger, O. 3D U-Net: learning dense volumetric segmentation from sparse annotation. Springer, City, 2016.Google Scholar
- Xiao, X., Lian, S., Luo, Z. and Li, S. Weighted res-unet for high-quality retina vessel segmentation. IEEE, City, 2018.Google ScholarCross Ref
- Zhou, Z., Rahman Siddiquee, M. M., Tajbakhsh, N. and Liang, J. Unet++: A nested u-net architecture for medical image segmentation. Springer, City, 2018.Google ScholarDigital Library
- Huang, H., Lin, L., Tong, R., Hu, H., Zhang, Q., Iwamoto, Y., Han, X., Chen, Y.-W. and Wu, J. Unet 3+: A full-scale connected unet for medical image segmentation. IEEE, City, 2020.Google ScholarCross Ref
- Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł. and Polosukhin, I. J. A. i. n. i. p. s. Attention is all you need, 30 (2017).Google Scholar
- Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G. and Gelly, S. J. a. p. a. An image is worth 16x16 words: Transformers for image recognition at scale (2020).Google Scholar
- Touvron, H., Cord, M., Douze, M., Massa, F., Sablayrolles, A. and Jégou, H. Training data-efficient image transformers & distillation through attention. PMLR, City, 2021.Google Scholar
- Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S. and Guo, B. Swin transformer: Hierarchical vision transformer using shifted windows. City, 2021.Google Scholar
- Chen, J., Lu, Y., Yu, Q., Luo, X., Adeli, E., Wang, Y., Lu, L., Yuille, A. L. and Zhou, Y. J. a. p. a. Transunet: Transformers make strong encoders for medical image segmentation (2021).Google Scholar
- Cao, H., Wang, Y., Chen, J., Jiang, D., Zhang, X., Tian, Q. and Wang, M. Swin-unet: Unet-like pure transformer for medical image segmentation. Springer, City, 2022.Google Scholar
- Vu, T.-H., Jain, H., Bucher, M., Cord, M. and Pérez, P. Advent: Adversarial entropy minimization for domain adaptation in semantic segmentation. City, 2019.Google Scholar
- Loshchilov, I. and Hutter, F. J. a. p. a. Sgdr: Stochastic gradient descent with warm restarts (2016).Google Scholar
- Tsai, A., Yezzi, A., Wells, W., Tempany, C., Tucker, D., Fan, A., Grimson, W. E. and Willsky, A. J. I. t. o. m. i. A shape-based approach to the segmentation of medical imagery using level sets, 22, 2 (2003), 137-154.Google Scholar
- Lieman-Sifry, J., Le, M., Lau, F., Sall, S. and Golden, D. FastVentricle: cardiac segmentation with ENet. Springer, City, 2017.Google Scholar
- Tran, P. V. J. a. p. a. A fully convolutional neural network for cardiac segmentation in short-axis MRI (2016).Google Scholar
- Valanarasu, J. M. J. and Patel, V. M. Unext: Mlp-based rapid medical image segmentation network. Springer, City, 2022.Google Scholar
- Devlin, J., Chang, M.-W., Lee, K. and Toutanova, K. J. a. p. a. Bert: Pre-training of deep bidirectional transformers for language understanding (2018).Google Scholar
- Parmar, N., Vaswani, A., Uszkoreit, J., Kaiser, L., Shazeer, N., Ku, A. and Tran, D. Image transformer. PMLR, City, 2018.Google Scholar
- Yang, X. and Tian, X. Transnunet: Using attention mechanism for whole heart segmentation. IEEE, City, 2022.Google ScholarCross Ref
- Lin, A., Chen, B., Xu, J., Zhang, Z., Lu, G., Zhang, D. J. I. T. o. I. and Measurement Ds-transunet: Dual swin transformer u-net for medical image segmentation, 71 (2022), 1-15.Google Scholar
- Pan, S., Liu, X., Xie, N. and Chong, Y. J. B. b. EG-TransUNet: a transformer-based U-Net with enhanced and guided models for biomedical image segmentation, 24, 1 (2023), 85.Google Scholar
- He, K., Zhang, X., Ren, S. and Sun, J. Deep residual learning for image recognition. City, 2016.Google Scholar
- Shannon, C. E. J. T. B. s. t. j. A mathematical theory of communication, 27, 3 (1948), 379-423.Google Scholar
- Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K. and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. Ieee, City, 2009.Google ScholarCross Ref
- Loshchilov, I. and Hutter, F. J. a. p. a. Sgdr: Stochastic gradient descent with warm restarts (2016).Google Scholar
- Asgari Taghanaki, S., Abhishek, K., Cohen, J. P., Cohen-Adad, J. and Hamarneh, G. J. A. I. R. Deep semantic segmentation of natural and medical images: a review, 54 (2021), 137-178.Google ScholarDigital Library
- Wang, R., Lei, T., Cui, R., Zhang, B., Meng, H. and Nandi, A. Medical image segmentation using deep learning: a survey. IET Image Process 16 (5): 1243‑1267. City, 2022.Google Scholar
Index Terms
- CardiacSegFormer: Transformer for Semantic Segmentation of Cardiac Images
Recommendations
Consistency Based Co-segmentation for Multi-view Cardiac MRI Using Vision Transformer
Statistical Atlases and Computational Models of the Heart. Multi-Disease, Multi-View, and Multi-Center Right Ventricular Segmentation in Cardiac MRI ChallengeAbstractSegmentation of cardiac structures in magnetic resonance imaging is essential to the diagnosis of many cardiovascular diseases. However, sometimes it is challenging to accurately define the right ventricle (RV) structure due to the complex ...
Semiautomatic Segmentation of CT Cardiac Images
ICICSE '13: Proceedings of the 2013 Seventh International Conference on Internet Computing for Engineering and ScienceIn order to complete the semi auto-segmentation of dual-source CT image of heart and extract the structure of heart accurately, propose a novel segmentation method of CT images based on graph cuts based active contour and anisotropies spreads ...
Cardiac image segmentation by random walks with dynamic shape constraint
The quantitative analysis of the left ventricle (LV) contractile function is one of the key steps in the assessment of cardiovascular disease. Such analysis greatly depends on the accurate delineation of LV boundary from cardiac sequences. However, ...
Comments