Gastrointestinal Polyp Segmentation with XAttUNet Based on the Attention-Driven Extended UNet Architecture

Öz

Timely and accurate detection of gastrointestinal polyps is crucial for the prevention and early diagnosis of colorectal cancer. This study introduces XAttUNet, a novel deep learning architecture developed for the polyp segmentation task. While preserving the basic structure of the UNet architecture, our model integrates attention mechanisms (CBAM), multi-head self-attention layers, and transformer structures to learn both local details and global contextual information more effectively. The proposed model was trained and evaluated on the Kvasir-SEG dataset. Experimental results demonstrate that XAttUNet provides significant improvements in metrics such as Dice, IoU, Precision, Recall, mAP, and Boundary-F1 compared to the classical UNet architecture. Furthermore, visual comparisons demonstrate that the model can capture polyp boundaries with greater precision and realism. In this respect, XAttUNet stands out as a solution that produces highly accurate and reliable results in the field of medical segmentation. This study aims to contribute to the development of deep learning-based polyp detection systems that can be used in clinical support systems.

Anahtar Kelimeler

• Segmentation
• Polyp segmentation
• Deep learning
• Attention mechanisms

Dikkat Odaklı Genişletilmiş UNet Mimarisine Dayalı XAttUNet ile Gastrointestinal Polip Segmentasyonu

Öz

Gastrointestinal poliplerin zamanında ve doğru bir şekilde tespiti, kolorektal kanserin önlenmesi ve erken tanısı açısından büyük önem taşımaktadır. Bu çalışmada, polip segmentasyonu görevine yönelik olarak geliştirilen yeni bir derin öğrenme mimarisi olan XAttUNet tanıtılmaktadır. Modelimiz, UNet mimarisinin temel yapısını korumakla birlikte, dikkat mekanizmalarını (CBAM), çoklu başlı self-attention katmanlarını ve transformer yapılarını entegre ederek, hem lokal detayları hem de küresel bağlamsal bilgiyi daha etkili bir biçimde öğrenebilmektedir. Önerilen model, Kvasir-SEG veri seti üzerinde eğitilmiş ve değerlendirilmiştir. Deneysel sonuçlar, XAttUNet’in klasik UNet mimarisine dayalı modellere kıyasla Dice, IoU, Precision, Recall, mAP ve Boundary-F1 gibi metriklerde anlamlı iyileşmeler sağladığını ortaya koymuştur. Ayrıca görsel karşılaştırmalar, modelin polip sınırlarını daha hassas ve gerçekçi şekilde yakalayabildiğini göstermektedir. Bu yönüyle XAttUNet, tıbbi segmentasyon alanında doğruluğu yüksek ve güvenilir sonuçlar üreten bir çözüm olarak öne çıkmaktadır. Yapılan bu çalışma, klinik destek sistemlerinde kullanılabilecek derin öğrenme temelli polip tespiti sistemlerinin geliştirilmesine katkı sunmayı amaçlamaktadır.

Anahtar Kelimeler

• Segmentasyon
• Polip segmentasyonu
• Derin öğrenme
• Dikkat mekanizmaları
• Transformer

Kaynakça

1. Sung, H. et al.: Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin, 71 (3), 2021, p. 209–249.
2. Zauber, A.G. et al.: Colonoscopic Polypectomy and Long-Term Prevention of Colorectal-Cancer Deaths. New England Journal of Medicine, 366 (8), 2012, p. 687–696.
3. Tajbakhsh, N. et al.: Automated polyp detection in colonoscopy videos using shape and context information. IEEE Trans Med Imaging, 35 (2), 2016, p. 630–644.
4. Khuc, T. et al.: Accuracy and Inter-observer Agreement Among Endoscopists for Visual Identification of Colorectal Polyps Using Endoscopy Images. Dig Dis Sci, 68 (2), 2023, p. 616–622.
5. Lee, J.Y. et al.: Real-time detection of colon polyps during colonoscopy using deep learning: systematic validation with four independent datasets. Sci Rep, 10 (1), 2020, p. 1–9.
6. Atık, I.: The Current Status And Future Vısıon Of Artıfıcıal Intellıgence Applıcatıons In Healthcare. Experimental and Applied Medical Science, 5 (2), 2024, p. 36–43.
7. Akçaman, M.N. et al.: Tıbbi Görüntü Bölütleme için MMIAU-Net Mimarisi Önerisi. Bilgisayar Bilimleri ve Mühendisliği Dergisi, 18 (1), 2025, p. 19–30.
8. Ronneberger, O. et al.: U-Net: Convolutional Networks for Biomedical Image Segmentation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 9351, 2015, p. 234–241.

9. Woo, S. et al.: CBAM: Convolutional block attention module. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11211 LNCS, 2018, p. 3–19.
10. Chen, J. et al.: TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation. 2021.
11. Cao, H. et al.: Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation. Lecture Notes in Computer Science, 13803 LNCS, 2021, p. 205–218.
12. Santone, A. et al.: A Method for Polyp Segmentation Through U-Net Network. Bioengineering 2025, Vol. 12, Page 236, 12 (3), 2025, p. 236.
13. Meng, W. et al.: AFC-Unet: Attention-fused full-scale CNN-transformer unet for medical image segmentation. Biomed Signal Process Control, 99, 2025, p. 106839.
14. Wang, H. et al.: Multilevel Attention Unet Segmentation Algorithm for Lung Cancer Based on CT Images. Computers, Materials & Continua, 78 (2), 2024, p. 1569–1589.
15. Chen, S. et al.: LD-UNet: A long-distance perceptual model for segmentation of blurred boundaries in medical images. Comput Biol Med, 171, 2024, p. 108120.
16. Salici, M., Olcer, U.E.: Impact of Transformer-Based Models in NLP: An In-Depth Study on BERT and GPT. 8th International Artificial Intelligence and Data Processing Symposium, IDAP 2024, 2024.
17. Zhou, Z. et al.: UNet++: A Nested U-Net Architecture for Medical Image Segmentation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11045 LNCS, 2018, p. 3–11.
18. Oktay, O. et al.: Attention U-Net: Learning Where to Look for the Pancreas. 2018.
19. Khan, A.R., Khan, A.: MaxViT-UNet: Multi-Axis Attention for Medical Image Segmentation. 2023.
20. Chowdary, G.J., Yin, Z.: Diffusion Transformer U-Net for Medical Image Segmentation. In: Medical Image Computing and Computer Assisted Intervention – MICCAI 2023 . Springer, Cham, 2023, p. 622–631.
21. Pham, T.H. et al.: SeUNet-Trans: A Simple yet Effective UNet-Transformer Model for Medical Image Segmentation. IEEE Access, 12, 2023, p. 122139–122154.
22. Rahman, M.M., Marculescu, R.: Medical Image Segmentation via Cascaded Attention Decoding. Proceedings - 2023 IEEE Winter Conference on Applications of Computer Vision, WACV 2023, 2023, p. 6211–6220.
23. Jha, D. et al.: Kvasir-SEG: A Segmented Polyp Dataset. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11962 LNCS, 2019, p. 451–462.
24. He, S. et al.: U-Netmer: U-Net meets Transformer for medical image segmentation. 2023.
25. Li, S. et al.: A novel recursive transformer-based U-Net architecture for enhanced multi-scale medical image segmentation. Comput Biol Med, 196, 2025, p. 110658.
26. Lin, L. et al.: GFANet: Global Feature Attention Network for Polyp Segmentation. Journal of Imaging Informatics in Medicine, 2025, p. 1–15.
27. Yu, J. et al.: Real-time colonoscopic detection and precise segmentation of colorectal polyps via PESNet. Front Oncol, 15, 2025, p. 1679826.
28. Adame, D. et al.: Topo-VM-UNetV2: Encoding Topology into Vision Mamba UNet for Polyp Segmentation. ArXiv, 2025.
29. Tang, R. et al.: A frequency attention-embedded network for polyp segmentation. Scientific Reports 2025 15:1, 15 (1), 2025, p. 4961.
30. Balasubaramanian, S. et al.: Attention-enhanced residual U-Net with histogram equalization for automated classification of gastrointestinal bleeding disorders. Biomed Signal Process Control, 111, 2026, p. 108371.
31. Akgöl, Y. et al.: Polyp segmentation with colonoscopic images: a study. Neural Comput Appl, 37 (17), 2025, p. 11311–11346.
32. Li, S. et al.: A survey of deep learning algorithms for colorectal polyp segmentation. Neurocomputing, 614, 2025, p. 128767.
33. Liu, X. et al.: Colorectal Polyp Segmentation Based on Deep Learning Methods: A Systematic Review. Journal of Imaging 2025, Vol. 11, Page 293, 11 (9), 2025, p. 293.
34. Li, S. et al.: A novel recursive transformer-based U-Net architecture for enhanced multi-scale medical image segmentation. Comput Biol Med, 196, 2025.