Pre-diagnosis of chronic and genetic diseases through facial images with deep learning methods

Chronic diseases generally require long-term treatment or care, and genetic diseases are health problems that can be transmitted from generation to generation through genes. Almost every disease negatively affects the quality of life of individuals. In recent years, it has become increasingly important to use deep learning algorithms effectively for the analysis of facial features, changing and developing anomalies in the face and disease diagnosis. In this study, the success of anomaly detection and disease classification from facial images with deep learning methods is investigated. This study evaluates the performance of the developed models compared to classical ESAs and aims to compare the success of different models. Transfer learning methods and different techniques were used to compare the performance of each model with other models. A classical ESA architecture and the ViT model are compared with binary disease success and performance. Multiple disease classification was performed with ESA architectures (Resnet50 and InceptionV3), visual transformers (ViT) and EfficientFormer. In the study, binary classification of all diseases, which we identified as four disease groups and one control group, was performed and the highest performance was obtained with the ViT model. Deep learning applications, especially the high success rate of the ViT model in facial diagnosis with small data sets, draw attention with the ability of newly developed models to produce effective results quickly and without the need for hybrid approaches. These developments show that these technologies can be used safely by making significant contributions to the prevention of human errors as well as reducing the patient burden of physicians in the medical field. In addition, the early diagnosis of some diseases without the need for interventional and radiological methods will increase the efficiency of health systems and provide significant savings in the health expenditures of countries. This will both reduce patient care costs and enable more efficient use of resources.