AI in medicine: can a selfie tell how long we might live?

In 2016, Megan Troutwine took a selfie in New York City and noticed one eyelid appeared droopy. After consulting a neurologist, magnetic resonance imaging (MRI) revealed a meningioma, a brain and spinal cord tumor. The photograph did not diagnose the tumour, but alerted Troutwine to a potential medical issue.

Scientists are now exploring artificial intelligence systems that analyze facial photos for signs of health, biological age, and survival prospects among cancer patients. These advances suggest personal photographs could serve as an additional source of medical information without specialized equipment or procedures.

However, identifying statistical patterns in facial features for reliable disease diagnosis or treatment guidance is still a long way off. Researchers are pursuing the ambitious goal of using algorithms to identify visual patterns containing health information beyond ordinary observation.

Photography has been used in medicine since the 19th century to document abnormalities and monitor disease-related changes. During the Covid-19 pandemic, smartphones with high-resolution cameras facilitated remote consultations and assessments, such as evaluating skin lesions suspected of being cancerous.

A study published in January in Annals of Plastic Surgery examined a remote assessment system involving 939 patients, aged with an average of 70.8 years. Doctors conducted telephone consultations based on patient-submitted photos and assessed the images for clarity, magnification, and accurate identification of the lesion's location.

The findings showed that 90.8% of patients were listed for surgery following remote assessment, but this figure does not confirm the diagnostic accuracy of photographs in detecting skin cancer. The images remain a supplementary tool alongside medical examinations and histological analysis to confirm diagnoses.