New Delhi: A new study explored how artificial intelligence can estimate a cancer patient’s survival chances using just one photograph of their face.
The technology was developed by researchers at Mass General Brigham, a nonprofit academic healthcare system based in Massachusetts, US. The findings, published in The Lancet Digital Health, highlight how facial image analysis could become a tool for assessing health risks and supporting medical decision-making in the future.
The AI system is designed to estimate a person’s biological age and assess survival outcomes among patients diagnosed with cancer.
Researchers trained the technology using more than 40 million facial images, along with photographs of over 30,000 cancer patients, many of whose medical outcomes were already known.
Hugo Aerts, a co-senior author of the study, highlighted the importance of predicting survival risk directly from facial images. He said this was the first demonstration of an AI model directly predicting cancer survival from a photograph.
“FAHR-FaceSurvival advances the field of facial health recognition AI a step further by learning survival risk directly from facial photographs of patients with cancer,” he said, as reported by The Guardian.
How do the two AI tools work?
Researchers developed two related AI tools, called FAHR-FaceAge and FAHR-FaceSurvival. Both were built from the same underlying model but perform different tasks.
FAHR-FaceAge estimates biological age, which refers to how old a person appears based on biological signs rather than their actual age. It can also help predict survival outcomes among cancer patients. This tool focuses on creases running from the sides of the nose down to the corners of the mouth, the forehead area just above the eye sockets, and the temples.
FAHR-FaceSurvival estimates a cancer patient’s survival risk directly from a facial photograph, without relying on biological age as an intermediate step. It evaluates the region below the eyes and around the nasal bridge.
Also Read: Rogue AI agent gives a fake tip on unsolved murder case. Police criticise Anthropic
How did the model perform across racial groups?
The study found that the model performed consistently across different age groups and racial backgrounds, showing strong correlations among White, Black, Asian, Hispanic, and other racial groups.
The researchers reported correlations of 0.86 for White individuals, 0.81 for Black individuals, 0.86 for Asian individuals, 0.91 for Hispanic individuals and 0.90 for other groups.
These results suggest that the model could estimate facial age across a range of populations. However, further research is needed to establish how reliably it performs in different clinical settings.
“FAHR-FaceAge showed generalisability across diverse age ranges and racial groups,” the study said.
The researchers also found links between lifestyle habits, including heavy drinking, drug use and smoking, and appearing biologically older in facial photographs analysed by AI.
“Another surprising finding was how consistent FAHR-FaceAge was across common real-world photo variations, including pose, lighting, facial expression and makeup,” Dr Raymond Mak, a radiation oncologist at Mass General Brigham, told The Guardian.
“It was also stable across most tested plastic surgery procedures, with only a small shift observed after rhinoplasty,” he added.
Although a single facial photograph could offer a simple, low-cost way to help doctors assess risks and plan personalised treatment, the technology is not yet ready to be used as a standalone tool for clinical decision-making.
(Edited by Maryam Hassan)