Comparative Analysis of Deep Learning Methods for Lesion Segmentation in Gastrointestinal Endoscopic Images

Gastrointestinal (GI) disorders present diagnostic obstacles, necessitating accurate imaging procedures. Conventional techniques for evaluating endoscopic images are mostly laborious and arbitrary. An innovative method is proposed to improve GI endoscopic image evaluation using three cuttingedge deep learning models, including Seg Net, U-Net, and Deep Lab, for lesion segmentation. Exploiting the recently published Kvasir-SEG dataset, which comprises segmentation masks and accompanied gastrointestinal polyp images, this work attempts to further the field of polyp analysis and detection research. The study aims to enhance diagnostic precision and optimize the analytical process by utilizing deep learning approaches and datasets that are annotated. It will eventually contribute for enhanced patient outcomes as well as healthcare productivity.