Detection and Classification of GI-Tract Anomalies from Endoscopic Images Using Deep Learning

Gastrointestinal tract in humans is affected by innumerable diseases. If not found earlier it may lead to cancer and death in the initial stages. Endoscopy is a well-established method for inspecting the human gastrointestinal tract. Some of the uneven morphologies may be missed by the physician during the examination. Deep learning techniques recent advances and high performance make them the best choice for computer-assisted diagnosis. Here, a deep learning model is suggested based on a deep CNN combined with a pre-trained model ResNet101 to detect and classify abnormalities in the GI tract. Disease detection in endoscopic images is the goal of the proposed research. The 8000-image KVASIR dataset, which is open to the public, serves as the foundation for the architecture. Our CNN approach improved and achieved an accuracy of 98.37%. The experiment demonstrates that the model can recognize at a higher level without the help of an individual.