Disease Detection in Gastrointestinal Endoscopic Images Using Deep Learning

Abstract

The diagnosis of gastrointestinal diseases by endoscopy is very important for early treatment. This work presents an automatic system for characterization of gastrointestinal endoscopic images based on deep learning methods. This system is aimed to discriminate Normal from multiple disease states like Polyps, Ulcerative Colitis, and Esophagitis. Image processing and data augmentation methods are used to improve the model's robustness. Different CNN (convolutional neural network) architectures like VGG16, InceptionV3 and ResNet50 are used and performance analyzed. The trained models can predict disease categories from input endoscopic images. For the performance analysis, standard measures such as accuracy, precision, recall, and F1-score are used to verify the performance of the proposed method. Overall, VGG16 provided a high classification accuracy of 91% compared to other models.