GI-ENSNet: A Hybrid Model For Segmenting Some Gastrointestinal Organs
Abstract
Gastrointestinal cancer is a leading cause of mortality, requiring highly
accurate treatment approaches such as radiotherapy. Accurate localization of organs such as the stomach, small intestine and large intestine on MRI scans is crucial to minimize radiation-induced damage to surrounding healthy tissues. However, manual segmentation is often time-consuming and prone to subjective errors. This study proposes a deep learning-based model named GI-ENSNet, which ensembles three architectures: UNet with EfficientNet-B2 as backbone, UNet++ with EfficientNet-B0 optimized using Stochastic Weight Averaging (SWA) and a modified UNet integrated with the Boundary Awareness Module (BAM). The outputs of these models are fused using a Soft Voting ensemble strategy, followed by post-processing with the Segment Anything Model (SAM) to refine segmentation masks. The model is trained on a dataset of 38,496 MRI images provided by the UW-Madison Cancer Center and achieves high segmentation performance with a Dice coefficient of 0.9255 and an IoU of 0.8964. The proposed approach significantly enhances segmentation accuracy and supports clinicians in delivering more precise radiotherapy while reducing processing time in clinical practice.
References
1. Ronneberger, O., Fischer, P., & Brox, T. (2015). Unet: Convolutional networks for bi-omedical image segmentation. In Medical image computing and computer-assisted inter-vention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 (pp. 234-241). Springer international publishing.
2. Tan, M., & Le, Q. (2019, May). Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning (pp. 6105-6114). PMLR.
3. Zhou, Z., Rahman Siddiquee, M. M., Tajbakhsh, N., & Liang, J. (2018). Unet++: A nested Unet architecture for medical image segmentation. In Deep learning in medical image analysis and multimodal learning for clinical decision support: 4th international workshop, DLMIA 2018, and 8th international workshop, ML-CDS 2018, held in con-junction with MICCAI 2018, Granada, Spain, September 20, 2018, proceedings 4 (pp. 3-11). Springer International Publishing.
4. Guo, H., Jin, J., & Liu, B. (2023). Stochastic weight averaging revisited. Applied Sci-ences, 13(5), 2935.
5. Sun, X., Shi, A., Huang, H., & Mayer, H. (2020). BAS $^{4} $ Net: Boundary-aware semi-supervised semantic segmentation network for very high resolution remote sensing images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 13, 5398-5413.
6. Kumari, S., Kumar, D., & Mittal, M. (2021). An ensemble approach for classification and prediction of diabetes mellitus using soft voting classifier. International Journal of Cognitive Computing in Engineering, 2, 40-46.
7. Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., ... & Girshick, R. (2023). Segment anything. In Proceedings of the IEEE/CVF international conference on computer vision (pp. 4015-4026).
8. Pogorelov, K., Randel, K. R., Griwodz, C., Eskeland, S. L., de Lange, T., Johansen, D., ... & Halvorsen, P. (2017, June). Kvasir: A multi-class image dataset for computer aided gastrointestinal disease detection. In Proceedings of the 8th ACM on Multimedia Sys-tems Conference (pp. 164-169).
9. Ronneberger, O., Fischer, P. and Brox, T., 2015. Unet: Convolutional networks for bi-omedical image segmentation. In Medical image computing and computer-assisted inter-vention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 (pp. 234-241). Springer International Publishing.
10. Yuan, Y., Li, D. and Meng, M.Q.H., 2017. Automatic polyp detection via a novel uni-fied bottom-up and top-down saliency approach. IEEE journal of biomedical and health informatics, 22(4), pp.1250-1260.
11. Jha, D., Smedsrud, P.H., Riegler, M.A., Johansen, D., De Lange, T., Halvorsen, P. and Johansen, H.D., 2019, December. Resunet++: An advanced architecture for medical im-age segmentation. In 2019 IEEE international symposium on multimedia (ISM) (pp. 225-2255). IEEE.
12. Poorneshwaran, J.M., Kumar, S.S., Ram, K., Joseph, J. and Sivaprakasam, M., 2019, Ju-ly. Polyp segmentation using generative adversarial network. In 2019 41St annual inter-national conference of the IEEE engineering in medicine and biology society (EMBC) (pp. 7201-7204). IEEE.
13. Tran, S.T., Cheng, C.H., Nguyen, T.T., Le, M.H. and Liu, D.G., 2021, January. Tmd-unet: Triple-unet with multi-scale input features and dense skip connection for medical image segmentation. In Healthcare (Vol. 9, No. 1, p. 54). MDPI.
14. Ghosh, S., Chaki, A. and Santosh, K.C., 2021. Improved UNet architecture with VGG-16 for brain tumor segmentation. Physical and Engineering Sciences in Medicine, 44(3), pp.703-712.
15. Punn, N.S. and Agarwal, S., 2022. Modality specific UNet variants for biomedical image segmentation: a survey. Artificial Intelligence Review, 55(7), pp.5845-5889.
16. Chou, A., Li, W. and Roman, E., 2022. GI tract image segmentation with UNet and mask R-CNN. Image Segmentation with UNet and Mask R-CNN. Available online: http://cs231n. stanford. edu/reports/2022/pdfs/164. pdf (accessed on 4 June 2023).
17. Zhou, H., Lou, Y., Xiong, J., Wang, Y. and Liu, Y., 2023. Improvement of Deep Learn-ing Model for Gastrointestinal Tract Segmentation Surgery. Frontiers in Computing and Intelligent Systems, 6(1), pp.103-106.
18. Sharma, N., Gupta, S., Reshan, M.S.A., Sulaiman, A., Alshahrani, H. and Shaikh, A., 2023. EfficientNetB0 cum FPN based semantic segmentation of gastrointestinal tract or-gans in MRI scans. Diagnostics, 13(14), p.2399.
19. Elsken, T., Metzen, J. H., & Hutter, F. (2019). Neural architecture search: A survey. Journal of Machine Learning Research, 20(55), 1-21.
20. Ye, R.; Wang, R.; Guo, Y.; Chen, L. SIA-Unet: A Unet with Sequence Information for Gastrointestinal Tract Segmentation. In Proceedings of the Pacific Rim International Conference on Artificial Intelligence, Shanghai, China, 10–13 November 2022; Spring-er: Cham, Switzerland, 2022; pp. 316–326
21. Chou, A.; Li, W.; Roman, E. GI Tract Image Segmentation with UNet and Mask R-CNN. CS231n: Deep Learning for Computer Vision, Stanford University. 2022. Available online: https://cs231n.stanford.edu/reports/2022/pdfs/164.pdf (accessed on 1 February 2025).
22. Qiu, Y. Upernet-Based Deep Learning Method for The Segmentation of Gastrointestinal Tract Images. In Proceedings of the 2023 8th International Conference on Multimedia and Image Processing, Tianjin, China, 21–23 April 2023; pp. 34–39.
23. Jiang, X.; Ding, Y.; Liu, M.; Wang, Y.; Li, Y.; Wu, Z. BiFTransNet: A unified and sim-ultaneous segmentation network for gastrointestinal images of CT & MRI. Comput. Biol. Med. 2023, 165, 107326
24. John, S.V.; Benifa, B. Automated Segmentation of Tracking Healthy Organs from Gas-trointestinal Tumor Images. In Proceedings of the International Conference on Frontiers of Intelligent Computing: Theory and Applications, Cardiff, UK, 11–12 April 2023; Springer Nature: Singapore, 2023; pp. 363–373.
25. Sharma, Neha, et al. "Encoder–Decoder Variant Analysis for Semantic Segmentation of Gastrointestinal Tract Using UW-Madison Dataset." Bioengineering 12.3 (2025): 309.