Artificial intelligence in modern medicine: a comprehensive review of technologies, applications, and future directions

Objectives We evaluated the transformative impact of AI modalities on clinical practice, distinguishing between retrospective algorithmic performance and prospective clinical effectiveness.Methods Comprehensive searches were executed [databases searched: January 2016 to October 2024] across PubMed/Medline, Scopus, and Web of Science, identifying peer-reviewed original research and clinical trials. Non-peer-reviewed materials and commercial web sources were strictly excluded.Results AI technologies—particularly deep learning (DL), natural language processing (NLP), and foundation models—demonstrate high diagnostic accuracy in retrospective cohorts across radiology, dermatology, and pathology. In precision therapeutics, AI accelerates drug discovery and facilitates multi-omics integration for individualized treatment planning. Furthermore, autonomous AI systems have proven cost-effective in nationwide diabetic retinopathy screening programs. However, significant gaps remain between in-silico success and real-world clinical utility. Primary barriers include the lack of prospective randomized trials, pervasive risks of algorithmic bias threatening health equity, the “black box” nature of deep neural networks, and stringent interoperability challenges.Conclusion AI serves as a potent computational catalyst for advancing personalized medicine. To transition from retrospective experimental success to standard-of-care clinical integration, future research endeavors must prioritize Explainable AI (XAI), federated learning frameworks for privacy preservation, and rigorous prospective clinical validation under standardized regulatory oversight.