Diabetic Prediction from Tongue Image

Abstract Elevated blood glucose levels are a hallmark of diabetes, a chronic metabolic disease that can cause serious health problems if left undiagnosed. Invasive blood tests, which are time-consuming and need clinical infrastructure, are the foundation of conventional diagnostic techniques. This study investigates a non-invasive diabetes prediction method based on tongue image analysis that was influenced by traditional Chinese medicine in order to overcome these drawbacks. Digital tongue photos are taken from people with and without diabetes, and they are pre-processed utilizing segmentation, normalization, and picture enhancement methods. Machine learning models, such as Support Vector Machines, Random Forests, and Convolutional Neural Networks, are used to extract and analyze pertinent aspects including color, texture, and shape.Because CNNs can automatically learn spatial properties from images, they perform better than the others. Promising results from an experimental evaluation using conventional metrics demonstrate the promise of tongue image-based analysis as an accessible and affordable tool for early diabetes screening, with potential incorporation into mobile health applications. Keywords: CNN, machine learning, non-invasive diagnosis, tongue image analysis, diabetes prediction.