Enhancing Polyp Segmentation Using Attention U-Net with CLAHE
Abstract
Colorectal cancer remains one of the leading causes of death worldwide, where early detection of polyps through colonoscopy plays a vital role in prevention. This study aims to enhance polyp segmentation performance by integrating Attention U-Net with Contrast Limited Adaptive Histogram Equalization (CLAHE) as a preprocessing technique. The proposed method was evaluated using two benchmark datasets, CVC-ClinicDB as the primary dataset and Kvasir-SEG for cross-domain testing. The model was trained using a combination of Binary Cross-Entropy and Dice losses, with a 70–15–15 split for training, validation, and testing. Experimental results show that applying CLAHE improves segmentation accuracy, achieving Dice and IoU scores of 0.84 and 0.76 on CVC-ClinicDB, and 0.62 and 0.50 on Kvasir-SEG, respectively. Statistical analysis using the Wilcoxon signed-rank test confirmed a significant difference between the baseline and enhanced models. These findings demonstrate that the integration of CLAHE with Attention U-Net effectively improves boundary detection and robustness against illumination variations across datasets, contributing to more accurate and reliable computer-aided diagnosis in colorectal cancer screening.
References
Bernal, J., Sánchez, F. J., Fernández-Esparrach, G., Gil, D., Rodríguez, C., & Vilariño, F. (2015). WM-DOVA maps for accurate polyp highlighting in colonoscopy: Validation vs. saliency maps from physicians. Computerized Medical Imaging and Graphics, 43, 99–111.
Cai, Y., Zhou, X., Li, X., & Wu, J. (2024). Enhanced reproducibility strategies for deep medical image segmentation models. Medical Image Analysis, 92, 103097.
Dong, Z., Zhang, Y., & Li, R. (2021). Attention-based encoder–decoder networks for small organ segmentation in endoscopic images. IEEE Access, 9, 1221–1234.
Fan, J., Lei, T., & Liu, Y. (2020). A review on data augmentation in medical image analysis. Journal of Imaging, 6(9), 1–17.
Fan, X., Wu, Y., & Zhu, Q. (2021). Improved attention gate mechanism for medical image segmentation. IEEE Transactions on Medical Imaging, 40(12), 3412–3424.
Fan, Y., Zhang, H., & Jiang, Z. (2023). Lightweight attention enhancement for polyp boundary detection. Pattern Recognition, 143, 109763.
He, X., Wang, H., & Li, P. (2025). CLAHE-driven enhancement for endoscopic image segmentation under low illumination. Expert Systems with Applications, 245, 123456.
Huynh, T., Nguyen, H., & Vu, T. (2024). Improved Attention U-Net for small-target segmentation in gastrointestinal imaging. Computer Methods and Programs in Biomedicine, 245, 107829.
Jiang, Y., Li, Z., & Chen, W. (2023). Attention-gated deep neural networks for precise medical image segmentation. Neural Networks, 165, 120–131.
Kim, S., Park, J., & Lee, H. (2023). Illumination-robust CNN models for endoscopic image analysis. Biomedical Signal Processing and Control, 83, 104642.
Li, M., Chen, X., & Zhao, Q. (2023). Improved contrast enhancement for endoscopic imaging using adaptive histogram equalization. Signal Processing: Image Communication, 115, 116197.
Liu, H., Chang, Z., & Luo, F. (2023). A comprehensive augmentation pipeline for medical segmentation networks. Artificial Intelligence in Medicine, 140, 102608.
Loshchilov, I., & Hutter, F. (2019). Decoupled weight decay regularization. International Conference on Learning Representations (ICLR).
Ma, Y., Xu, S., & Lin, J. (2024). Benchmarking evaluation metrics for medical image segmentation: A review and experimental study. Medical Image Analysis, 91, 103123.
Milletari, F., Navab, N., & Ahmadi, S.-A. (2016). V-Net: Fully convolutional neural networks for volumetric medical image segmentation. 4th International Conference on 3D Vision (3DV).
Namazi, H., & Shih, F. (2024). Adaptive CLAHE for robust medical image preprocessing under illumination variation. IEEE Transactions on Biomedical Engineering, 71(2), 551–562.
Nie, Y., Zhou, H., & Wang, S. (2024). A comprehensive review of colorectal cancer imaging and early polyp detection. Frontiers in Oncology, 14, 12984.
Oktay, O., Schlemper, J., Le Folgoc, L., et al. (2018). Attention U-Net: Learning where to look for the pancreas. arXiv preprint arXiv:1804.03999. https://arxiv.org/abs/1804.03999
Pogorelov, K., Riegler, M., Halvorsen, P., & de Lange, T. (2017). Kvasir: A multi-class dataset for gastrointestinal image analysis. Proceedings of MMSys, 164–169.
Santone, A., Brunese, L., & Mercaldo, F. (2025). Deep CNN advancements in computer-aided colorectal cancer diagnosis. Artificial Intelligence in Medicine, 150, 102918.
Taha, A. A., & Hanbury, A. (2015). Metrics for evaluating 3D medical image segmentation: Analysis, selection, and tool. BMC Medical Imaging, 15, 29.
Tang, L., Huang, Y., & Sun, Z. (2024). Adaptive histogram equalization for enhanced endoscopic image contrast. Image and Vision Computing, 142, 104817.
Wang, X., Liu, J., & Chen, Q. (2025). Robust polyp detection under varied illumination using deep contrast-aware CNNs. Pattern Recognition, 150, 110201.
WHO. (2024). Colorectal cancer statistics 2024. World Health Organization. https://www.who.int
Wu, D., Zhang, K., & Li, F. (2024). Deep attention-based architectures for colorectal polyp segmentation: A systematic review. Biomedical Signal Processing and Control, 89, 105573.
Yoshimi, T., Nakajima, K., & Mori, Y. (2024). CLAHE-based preprocessing for enhanced gastrointestinal endoscopic imaging. Endoscopy International Open, 12(3), E310–E318.
