EfficientNet-B4 Based Deep Learning Approach for Accurate Detection of Gastrointestinal Disorders

This innovative approach detects for ulcerative colitis, polyps, esophagitis and normal using a revised EfficientNet-B4 model. The dataset contains 6000 endoscopic images. They belong to four classes and are evenly distributed in each class: ulcerative colitis, polyps, esophagitis, and normal. The modified EfficientNet-B4 model, which balances computational efficiency with high accuracy, was chosen due to its versatility. Several modifications resulting in better performance were done to that model: fine-tuning of hyperparameters, application of advanced data augmentation techniques. After training with these modified settings, the model achieved a remarkable accuracy of 99.44% on the test set. This unusually high accuracy simply confirms that the modified EfficientNet-B4 model is a solid tool for accurate and efficient detection of varied gastrointestinal conditions we examined. The results of this study show that the method we propose could greatly benefit early diagnosis and treatment of these diseases resulting in Increasing life expectancy. This would in turn help to improve patient outcomes and reduce health care costs.