Medical imaging foundation models are reshaping how artificial intelligence (AI) may interpret scans, combine clinical information, and support multiple tasks from a shared computational base. Rather than training a separate model for every disease or imaging problem, the approach uses large-scale pre-training to encode reusable knowledge into generalizable representations, which can then be adapted for classification, segmentation, detection, report generation, medical question answering, and prognosis. A new review maps the field's technical routes and emerging applications while warning that impressive benchmark scores do not yet prove clinical benefit. It argues that progress should be judged by generalizability, reliability, workflow value, safety, and accountability--not simply by model size or the number of tasks covered.