Low-Dose CT in the Age of Artificial Intelligence: Techniques, Evidence, and Clinical Impact

Low-dose computed tomography (LDCT) is essential for reducing potential radiation-associated risks, but increased noise and artifacts can compromise diagnostic confidence. This review examines how AI is reshaping LDCT by improving reconstruction, denoising, protocol optimization, and workflow integration. It outlines the evolution from FBP to IR and then to deep learning methods that better preserve anatomical detail while reducing noise. The review also considers emerging applications in photon-counting CT (PCCT) and discusses clinical use in screening, pediatric imaging, oncology, cardiac imaging, and emergency care. These developments may support safer imaging, improve consistency across studies, and reduce the need for repeat examinations. However, important challenges remain, including limited generalizability across scanners and patient populations, the potential loss of subtle findings, changes in quantitative imaging features, and the need for careful validation before widespread clinical adoption. Further research should focus on multicenter validation, task-based performance assessment, and integration into routine clinical workflows.