Gastrointestinal disease prediction using transfer learning

According to the Global Cancer Observatory's 2018 estimations, gastrointestinal cancer is the third most common cause of cancer deaths and the fifth most common cause of being diagnosed with cancer worldwide. Diagnosis is crucial, and gastroscopy is used to detect stomach cancer early to enhance patient survival. This study proposed a Deep Learning based Computer-Aided Diagnostics (CADx) method to identify gastroscopy disease. The dataset is trained with an adversarial training technique. The proposed approach uses the deep convolution neural network based on VGG16, VGG19, and InceptionV3. The dataset is trained with an adversarial training technique. The Inception V3 has the best accuracy among them, with an accuracy in the training of 94.44% and a validation accuracy of 84.75%, with a minor loss compared to VGG16 and VGG19 models.