GI TRACE IMAGE SEGMENTATION WITH UNET KERAS (KEY IJP************299)

• J Ravi

Abstract

Advancements in medical imaging have paved the way for innovative solutions in diagnosing and treating gastrointestinal (GI) tract-related disorders. Image segmentation, a crucial step in medical image analysis, plays a vital role in identifying and delineating specific regions of interest within these images. This project focuses on the development and implementation of a deep learning based approach for GI tract image segmentation, utilizing the U-Net architecture with the Keras framework. The U-Net architecture has proven to be highly effective in biomedical image segmentation tasks due to its ability to capture both local and global features. The proposed model is trained on a curated dataset of GI tract images, encompassing a diverse range of anatomical structures and pathologies. The training process involves optimizing the model using a combination of loss functions tailored for segmentation tasks, such as dice coefficient loss and binary cross-entropy. The paper aims to achieve accurate and reliable segmentation results, enabling clinicians to obtain detailed insights into the structure and composition of the GI tract. The system's performance is evaluated through quantitative metrics, including precision, loss, and Dice similarity coefficient, and compared against existing segmentation methods.