AI models in the early diagnosis of diseases

Abstract

The fast evolution of artificial intelligence (AI) technology has brought about drastic changes in the field of medicine. In particular, when it comes to early disease diagnostics, systems based on machine learning, deep learning and large language models present an accurate and swift alternative to the conventional diagnostic methods available. AI-based systems are capable of analysing enormous volumes of multimodal medical data, which cannot be managed by humans, such as data from various types of medical imaging, genome sequences, electronic health records, and real-time physiological parameters. This article presents the application of AI-based systems in the areas of oncology, cardiology, neurology and infectious diseases. AI diagnostics is shown to provide diagnostic accuracy equal to or surpassing that achieved by doctors. However, there are still many technical, ethical and legal issues associated with the use of such systems in practice, such as biased algorithms, transparency issues, and problems with privacy and data regulation. The development of explainable artificial intelligence (XAI) and federated learning methods is bringing solutions to some of these problems and making the transition to clinical use possible.