Paper Title
Detection and Segmentation of Colon Polyps: A Deep Learning Approach with U-Net and ACSNet

Abstract
The early identification of colon polyps is vital to prevent colorectal cancer, a leading health issue globally. This study employs a deep learning approach using a U-Net model integrating the ACSNet module, specifically tailored to segment polyps in colonoscopy images. Leveraging the annotated Kvasir-SEG dataset, which provides diverse 1000 images with labeled regions, this model undergoes a structured workflow encompassing data preprocessing, augmentation, and specialized training for optimal segmentation. Incorporating ACSNet modules into U-Net enhances feature extraction through attention layers, improving the model’s ability to distinguish polyps from surrounding colon tissue, thus achieving more precise segmentation. It achieves 85% accuracy in identifying polyps and 97% accuracy in segmentation, indicating robust and reliable performance. This system enhances diagnostic workflows, supporting both real-time and offline analysis, and offers clinicians a valuable tool to improve patient outcomes. Results emphasize Integrated U-Net’s capability as a practical solution for diagnostic imaging in colorectal cancer prevention, potentially transforming automated medical image segmentation in clinical settings. Keywords - Colon Polyp Detection, Medical Image Segmentation, U-Net Architecture, ACSNet, Kvasir-SEG Dataset, Deep Learning in Medical Imaging