Recalibration Parallel Half U-Net for Polyp Segmentation

Polyp segmentation in colonoscopy images is a critical task for the early detection and prevention of colorectal cancer, which remains one of the leading causes of cancer- related deaths worldwide. Accurate segmentation of polyps is essential to assist medical professionals in identifying and removing precancerous lesions during routine screening procedures. In this study, we propose the Recalibration Parallel Half U-Net, a novel deep learning architecture based on the Half U-Net framework, designed to enhance segmentation performance through a dual parallel encoder structure. This recalibration mechanism effectively captures multi-scale features and improves the model’s ability to differentiate between polyp and non-polyp regions. Experimental results on the widely used CVC-Clinic dataset demonstrate that our method significantly outperforms traditional U-Net and Half U-Net models, achieving a mean F1-score of 93.63% and a mean Intersection over Union (IoU) of 88.47%. These performance gains highlight the robustness and accuracy of our proposed approach, indicating its potential for integration into clinical workflows to support more reliable and efficient polyp detection in colonoscopy procedures.