Abstract:
Endoscopy is the most widely adhered medical procedure used to examine the gastrointestinal tract of a person. Accurate pathology detection during the endoscopic procedure is crucial as misidentifications or miss rates could reduce the chance of survival for the patient. After the successful collaboration of artificial intelligence with medicine, researchers around the world have tried different techniques in using this for gastroenterology. Our study demonstrates an extensive survey on existing pathology detection methodologies in endoscopic images using the publicly available datasets. The paper also discusses the content of the recently released datasets, preprocessing techniques tried on these datasets and how they affected the performance of the machine learning models. Furthermore, this study discusses how changing architectures of convolutional neural networks could affect the accuracy of models in relation to different datasets. Finally, the paper presents the results of each reviewed literature along with a brief discussion on the gaps that were identified.
