Artificial intelligence is increasingly being explored as a way to expand imaging capacity in areas where radiologists and other imaging professionals are in short supply. But one of the most important questions for global health is whether an AI model developed using data from one population will perform equally well in another.

Recent research into AI-driven cervical cancer screening provides an interesting example. Researchers evaluated an AI model across datasets from different countries and found meaningful differences in performance between populations. The findings highlight an important consideration for the global deployment of AI: models need to be evaluated in the populations and healthcare environments where they will actually be used.

For low- and middle-income countries, this issue is particularly important. AI has the potential to help extend the reach of imaging services, but successful implementation requires representative data, local validation, appropriate infrastructure, trained personnel, and ongoing quality monitoring.

As the ISR continues to support the responsible use of AI in global imaging, questions of equity, local data, validation, and clinical governance will remain central to the conversation.