Since the concept of artificial intelligence （AI） was proposed in 1956， medicine has been one of its core application fields. At present， AI technology has run through the whole diagnosis and treatment process， and has been extended to innovative scenarios such as drug research and development， surgical robots， and clinical trial optimization. Scenario application is the backbone of the technical system. Multimodal data fusion integrates heterogeneous data such as images， medical records， and genes， and federated learning realizes cross-institutional privacy protection and sharing. Deep learning achieved more than 90% sensitivity in imaging diagnosis for lung nodule detection， while generative AI accelerates drug molecule design. The core applications cover four major areas field： AI is more accurate than professional doctors in breast cancer and diabetic retinopathy screening； robotics shortens hospital stays and improves spinal screw placement accuracy； AI shortens the discovery cycle of drug targets； machine learning improves the efficiency of subject screening and enables real-time data monitoring. The application of AI in the medical field is first constrained by data quality and algorithm bias， and the “black box” characteristics of AI models and the ambiguity of responsibility attribution are the core obstacles to clinical implementation. This paper analyzes key technological breakthroughs and typical cases， discusses the application scenarios and challenges of AI in medicine， and aims to provide a reference for the future development of medical intelligence.