This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
This paper presents a lightweight and efficient framework for rapid and accurate detection of colonic polyps, which play an important role in the early diagnosis and prevention of colon cancer. We propose a method to filter out noisy data using the LOF algorithm and detect colonic polyps using the YOLO-v11n deep learning model. We conducted experiments using five public datasets, including CVC-ColonDB, CVC-ClinicDB, Kvasir-SEG, ETIS, and EndoScene, and improved the robustness and generalization performance of the model through 5-fold cross-validation and LOF-based outlier removal. As a result, we achieved high accuracy (precision 95.83%, recall 91.85%, F1-score 93.48%, mAP@0.5 96.48%, mAP@0.5 :0.95 77.75%), showing improved accuracy and efficiency than the existing YOLO-based method.
Takeaways, Limitations
•
Takeaways:
◦
We propose the possibility of building an efficient and accurate colon polyp detection framework by combining the LOF algorithm and the YOLO-v11n model.
◦
Emphasizes the importance of data preprocessing and model efficiency in medical image analysis.
◦
Presenting the possibility of use in developing a real-time colonoscopy support system.
•
Limitations:
◦
Possible limitations in generalization performance due to restrictions on the type and scope of the dataset used.
◦
Lack of validation in real clinical settings.
◦
Further research is needed to optimize the parameters (number of neighbors, contamination ratio) of the LOF algorithm.
