US doctors' AI tool can predict a cancer patient's survival chances from a single photo of their face

FaceSurvival, developed at Mass General Brigham in Boston, detects clues from features such as under-eye bags and the nose bridge. The model was trained on over 40 million faces before being tested against photographs from a cohort exceeding 34,000 cancer patients. Those whose apparent age outstripped their actual age by five or more years faced a 21 per cent greater likelihood of death. The research appears in the Lancet Digital Health.

More than half of patients flagged highest-risk died within 90 days, against two per cent lowest-risk.

Smoking aged faces most: current smokers looked older than ex-smokers, who looked older than non-smokers.

Professor Hugo Aerts called it 'the first demonstration' of predicting cancer survival from a photograph.