Abstract
The evolution of medicine has progressed through distinct ages: from empirical observation and evidence-based practice to the current era of precision medicine. However, traditional healthcare paradigms remain constrained by data fragmentation, scalability limits, and the overwhelming complexity of multi-omics integration. In the rapid explosion of artificial intelligence (AI), a transformative paradigm is emerging. This review introduces the concept of "Virtual Medicine", which is defined as a comprehensive ecosystem of AI-empowered medical practice that transcends physical limitations. This review systematically summarizes the technological foundations, historical evolution, and core applications of AI in medicine, including electronic health records (EHRs) analysis, medical imaging, multimodal diagnostics, drug discovery, precision oncology, intelligent surgery, and clinical decision support systems. It further highlights the role of medical AI in health management, public health surveillance, and healthcare delivery in resource-limited settings. Special attention is given to the transformative emergence of large language models (LLMs), such as medical large language models (MedLLM) and generative pre-trained transformer (GPT) architectures, emphasizing their potential to revolutionize virtual medical interaction, clinical reasoning, and documentation. Despite these advances, significant challenges remain regarding model transparency, data bias, fairness, and patient privacy. Overcoming these limitations necessitates standardized evaluation frameworks, interpretable algorithm designs, and strengthened privacy protections. Ultimately, these efforts aim to foster a trustworthy and equitable future for virtual medicine.