Abstract

In gastrointestinal endoscopy, polyp identification is an important step in the fight against colorectal cancer. Using the cutting-edge BASNet model—a deep learning architecture for picture segmentation—we provide a web-based polyp identification method in this study. Endoscopy pictures may be uploaded to the system and evaluated by the BASNet model to detect and highlight possible polyps. In order to correctly separate polyps from endoscopic pictures, the suggested method makes use of the BASNet model. This model combines a prediction module with a residual refinement module. The segmentation is refined to provide a fine label map by the residual refinement module, after the prediction module has provided a coarse one. In order to reliably and accurately forecast where polyps are located within the pictures, the system makes use of these components. Model loading for rapid inference, real-time prediction of polyp masks, and image preprocessing to guarantee consistency in input data are key components of the system. In order to help medical professionals intervene quickly and improve patient outcomes, the technology gives users instant feedback on polyp diagnosis.