Artificial Intelligence in Radiology: Current Applications, Clinical Challenges, and Future Perspectives

Abstract

Artificial intelligence (AI) is rapidly transforming radiology by supporting medical image analysis, diagnosis, quantitative assessment, workflow optimization, and reporting. Advances from conventional computer-aided diagnosis to machine learning, deep learning, transformers, foundation models, generative AI, and multimodal systems have expanded AI applications across CT, MRI, chest radiography, mammography, ultrasound, PET, and SPECT. However, clinical implementation remains challenging because of dataset bias, limited generalizability, false predictions, explainability, automation bias, privacy, cybersecurity, and regulatory concerns. Generative and multimodal AI offer new opportunities for integrating imaging with clinical information but introduce additional risks, including hallucination and incorrect reasoning. This narrative review summarizes the major applications, benefits, limitations, validation requirements, workflow integration, ethical considerations, and future directions of AI in radiology. Current evidence supports a human-AI partnership in which AI enhances radiologists’ capabilities while radiologists retain clinical judgment and final decision-making responsibility.