Multi-Class Gastrointestinal Images Classification Using EfficientNet-B0 CNN Model

Abstract

Many diseases and cancerous cells can be detected using images taken by gastroenterology specialists. Accurate and rapid detection of gastroenterological diseases is very important for the treatment processes to be applied and for the patient’s recovery. In this study, a data set containing data from 8 different diseases (Esophagitis, Dyed and Lifted Polyps, Dyed Resection Margins, Cecum, Pylorus, Z-line, Polyps, Ulcerative colitis) was used. A deep learning network was trained using the EfficientNet architecture and the test results were given in the study. In addition, comparisons were made with other studies using the same data set and the same parameters in the literature. Studies have shown that gastrological images can be successfully classified with an accuracy of 0.935. Class-based classification results are also shared in detail for 8 diseases in the results section of the study. The results showed that the trained architecture would contribute to minimizing human error in disease detection.