Identification of Gastric Abnormalities with Synchronous Biopsy Site Localization from Endoscope Imaging in Gastro-Intestinal Tract using Deep Learning Technique

One of the most prevalent malignant tumors with a dismal prognosis is gastric cancer (GC). Although pathologic verification and CT scanning are recommended for additional therapy, endoscopic examination is mostly used for early detection. One of the worst diseases with a dismal prognosis is still gastric cancer. Artificial intelligence (AI) assistance technologies have a great opportunity to reduce workload and improve diagnostic accuracy due to the global pathologist shortage. The majority of gastric cancers (GCs) exhibit genetic instability, either as chromosomal or microsatellite instability, which is thought to be a precursor to gastric carcinogenesis. The features of each subtype are better understood because of the new classification of gastric malignancies on the basis of histologic features, genotypes, and molecular phenotypes, which also improves early diagnosis, prevention, and treatment. As a result, the suggested system offers a practical way to more accurately forecast the categories of gastrointestinal problems. In this research, the types of diseases will be identified by the use of Deep Learning algorithms, such as CNN. The datasets will be gathered from Kaggle and preprocessed using several methods, including adaptive median filtering. The UNet algorithm is used to segment the pictures after preprocessing. After that, a deep learning algorithm will be used to train these datasets, and a model file will be produced. It is able to accurately identify gastrointestinal anomalies such as ulcers, funduses, healthy tissues, and esophagitis when given an input image for prediction. As a result, this approach contributes to a more accurate and efficient detection of gastrointestinal problems than current models.