Artificial intelligence (AI) can estimate a cancer patient’s chances of survival by analysing a single photo of their face. The new technology can also work out the patient’s biological age compared with their actual age. Doctors say the AI-based system could in future play an important part in choosing treatment and assessing risk.
The AI tool was developed by experts at Mass General Brigham, a non-profit integrated academic healthcare system in the US state of Massachusetts. It was trained on more than 40 million photos of faces, covering people of different ethnic groups and ages.
The findings were published in the journal The Lancet Digital Health, the British newspaper The Guardian reported on Monday (19 October).
The researchers built two separate tools on the same model. FAHR-FaceAge estimates a patient’s biological age from a photo of their face and also predicts their chances of surviving cancer. FAHR-FaceSurvival, by contrast, works out survival chances directly from the face photo without relying on an estimate of biological age.
Although both tools come from the same AI model, they pick up different and complementary information about a patient’s health, and they look at different parts of the face to judge age and risk of death.
FAHR-FaceAge analyses the folds running from the sides of the nose to the corners of the mouth, the part of the forehead just above the eye sockets and the areas at the sides of the forehead. FAHR-FaceSurvival focuses on the area under the eyes and around the top of the nose.
Hugo Aerts, a co-senior author of the study, said face-analysis AI had moved a step further by learning to estimate survival directly from photos of cancer patients. To their knowledge, he said, it is the first example of an AI model directly estimating a cancer patient’s chances of survival from an ordinary photo.
The technology was tested on cancer patients receiving radiotherapy in the United States. Patients whom FAHR-FaceAge judged to look at least five years older than their actual age had a 21 percent higher risk of death than other patients over an average follow-up of four years. Those who looked at least five years younger had a 17 percent lower risk.
Aerts said the key advance is learning general facial features from more than 40 million images. That knowledge can then be used to measure various health markers, such as biological age and survival chances. Rather than building a new model from scratch for each task, the approach makes it possible to build systems that measure several health markers.
The researchers say biological age as estimated by AI is closely linked to a person’s physical health and may indicate their ability to tolerate certain treatments or survive afterwards.
In future the technology could also help gauge the risk of death and complications after major surgery or other cancer treatment, in particular by helping doctors assess a patient’s strength and frailty before intensive treatment.
Dr Raymond Mak, a radiation oncologist at Mass General Brigham, said that as the population ages, there is a growing need to assess a patient’s strength and frailty accurately before intensive treatment such as radiotherapy, surgery or drugs that act on the whole body.
Their aim, he said, is to make health assessment from face photos more measurable, consistent and accessible. It could help tailor treatment decisions to a patient’s circumstances, but not as a substitute for a doctor’s own judgement.
বাংলায় মূল প্রতিবেদন পড়ুন · Read the original Bengali report