Keywords
Medical imaging
YOLOv5n
CARAFE upsampling operator
ContextGuidedC3 module
How to Cite
Abstract
Gastrointestinal polyps are common lesions in the digestive system, and some lesions carry a risk of malignant transformation. Timely and accurate detection is of great clinical significance. However, due to factors such as weak polyp texture, low contrast, and complex backgrounds, traditional endoscopic examinations still have a certain detection missed rate. To address this, this paper proposes an improved network for gastrointestinal polyp detection based on the lightweight object detection model YOLOv5n. The proposed method introduces the ContextGuidedC3 module to enhance multi-scale contextual information modeling. Additionally, the CARAFE (Content-Aware Re-Assembly of Feature Elements) operator is used for content-aware upsampling, improving the reconstruction quality of high-resolution features and the expression of small target edge details. Experiments were conducted on the polyps dataset and compared with lightweight models such as YOLOv5n, YOLOv8n, YOLOv10n, and YOLOv11n. The results show that, compared to the original YOLOv5n, the proposed method improved the Recall from 85.7% to 90.3%, mAP@0.5 from 92.4% to 94.2%, and mAP@0.5-0.95 from 67.3% to 69.6%. The method also demonstrates better stability and robustness in terms of confusion matrix, visualized detection results, and PR curves. This indicates that the proposed method effectively improves the detection accuracy and small target recognition capability of gastrointestinal polyps while maintaining low computational overhead, and has potential clinical application value.
References
[1] Jahn, B., Bundo, M., Arvandi, M., Schaffner, M., Todorovic, J., Sroczynski, G., ... & Siebert, U. (2025). One in three adenomas could be missed by white-light colonoscopy–findings from a systematic re-view and meta-analysis. BMC gastroenter-ology, 25(1), 170.
[2] Taghiakbari, M., Mori, Y., & von Renteln, D. (2021). Artificial intelligence-assisted colonoscopy: A review of current state of practice and research. World journal of gas-troenterology, 27(47), 8103.
[3] Nie, M. Y., An, X. W., Xing, Y. C., Wang, Z., Wang, Y. Q., & Lü, J. Q. (2024). Artifi-cial intelligence algorithms for real-time de-tection of colorectal polyps during colonos-copy: a review. American Journal of Cancer Research, 14(11), 5456.
[4] Dougherty, K. E., Melkonian, V. J., & Montenegro, G. A. (2021). Artificial intel-ligence in polyp detection-where are we and where are we headed?. Artificial Intel-ligence in Gastrointestinal Endoscopy, 2(6), 211-219.
[5] Rostami, G., Hosseini Berneti, S. H., Habibzadeh, N., & Bazir, M. (2025). From Colon to Uterus: Potential of YOLOv7 for Real-Time Polyp Detection in Hysterosco-py. InfoScience Trends, 2(4), 48-57.
[6] Haider, Z., Azar, A. T., Fati, S. M., & Ibra-heem, I. K. (2025, February). Deep Learn-ing and AI for Superior Colorectal Polyp Detection with YOLOv9 Variants. In 2025 8th International Conference on Data Sci-ence and Machine Learning Applications (CDMA) (pp. 150-155). IEEE.
[7] Sun, X., Ma, J., & Li, Y. (2025). Efficient polyp detection algorithm based on deep learning. Scandinavian Journal of Gastro-enterology, 60(6), 502-515.
[8] Keshavarz, H., Ansari, Z., Abootalebian, H., Sabet, B., & Momenzadeh, M. (2025). Introducing a Deep Neural Network Model with Practical Implementation for Polyp Detection in Colonoscopy Videos. Journal of Medical Signals & Sensors, 15(6), 17.
[9] Viet, H. D., Nguyen, T. T., Lam, H. N., Nguyen, B. P., Vu, T. Q., Nguyen, H. M., ... & Nguyen, T. T. (2025). Validation of YOLOv8 algorithm in detecting colon polyps in endoscopy videos. Journal of Medical Artificial Intelligence, 8, 35.
[10] Chen, S., Chen, T., Chen, J., Chen, H., Chen, Y., & Zhang, J. (2025, June). YOLO-MF: A Multi-Branch Feature Sensing and Fine-Grained Calibration Network for Gas-troenteroscopic Polyp Detection. In 2025 5th International Conference on Computer Graphics, Image and Virtualization (IC-CGIV) (pp. 78-82). IEEE.
[11] Jocher, G., Stoken, A., Chaurasia, A., Bo-rovec, J., Kwon, Y., Michael, K., ... & Thanh Minh, M. (2021). ultralytics/yolov5: v6. 0-YOLOv5n'Nano'models, Roboflow integration, TensorFlow export, OpenCV DNN support. Zenodo.
[12] Wu, T., Tang, S., Zhang, R., Cao, J., & Zhang, Y. (2020). CGNet: A light-weight context guided network for semantic seg-mentation. IEEE Transactions on Image Processing, 30, 1169-1179.
[13] Wang, J., Chen, K., Xu, R., Liu, Z., Loy, C. C., & Lin, D. (2019). Carafe: Content-aware reassembly of features. In Proceed-ings of the IEEE/CVF international confer-ence on computer vision (pp. 3007-3016).
[14] Silva, Juan, et al. "Toward embedded detec-tion of polyps in wce images for early di-agnosis of colorectal cancer." International journal of computer assisted radiology and surgery 9.2 (2014): 283-293.
[15] Jha, Debesh, et al. "Kvasir-seg: A segment-ed polyp dataset." International conference on multimedia modeling. Cham: Springer International Publishing, 2019.
[16] Ming Chen, Tingting Chen, Yukang Lou, Yan Li, Jiyang Yu; Remote sensing image detection method based on context-aware mechanism and transformer architecture. AIP Advances 1 July 2025; 15 (7): 075326. https://doi.org/10.1063/5.0283520
[17] Wang, Z., Hua, Z., Wen, Y., Zhang, S., Xu, X., & Song, H. (2024). E-YOLO: Recogni-tion of estrus cow based on improved YOLOv8n model. Expert Systems with Applications, 238, 122212.
[18] Wang, A., Chen, H., Liu, L., Chen, K., Lin, Z., & Han, J. (2024). Yolov10: Real-time end-to-end object detection. Advances in Neural Information Processing Systems, 37, 107984-108011.
[19] Khanam, R., & Hussain, M. (2024). Yolov11: An overview of the key architec-tural enhancements. arXiv preprint arXiv:2410.17725.
This work is licensed under a Creative Commons Attribution 4.0 International License.
Copyright (c) 2025 Frontiers in Preventive Medicine