A photo of a face is enough for an artificial intelligence algorithm to assess the prognosis of a cancer patient, according to specialists at Mass General Brigham in Boston. Such an analysis can be helpful, but it will not replace specialized tests of a patient’s health.
AI models are successfully used to estimate a person’s biological age, which does not always match chronological age. They analyze various features of the face, particularly the temples, forehead, the area around the eyes and the nasolabial fold (the line in the skin running from the sides of the nose to the corners of the mouth).
One AI model in this area, called FAHR-FaceAge, has already been trained on more than 40 million photos of human faces. It takes just one selfie to estimate biological age.
A new model called FAHR-FaceSurvival is more specialized. It predicts the prognosis of a cancer patient from a photo of the face. It differs in that it assesses other regions of the face, mainly the bridge of the nose and the area under the eyes.
Specialists from Mass General Brigham in Boston reported in The Lancet Digital Health that among 122 cancer patients whom FAHR-FaceSurvival rated as being at high risk of death, more than half (52.5%) died within 90 days. In the comparison group of 975 people with low-risk cancer, only 2% died in the same period.
Survival predictions from the FAHR-FaceAge model were also compared. Of 1,060 people whose biological age was 20 years above their chronological age, 21% died within 90 days. Among 4,195 people whose biological age was at least 10 years younger than their chronological age, 6% died within three months.
The U.S. researchers say the AI accurately assesses biological age and prognosis from a facial photo even in people who were wearing makeup, as well as those who had undergone cosmetic procedures, such as Botox.
However, analyzing a facial photo of cancer patients is not meant to replace specialized tests, particularly molecular tests of cancer cells that assess how aggressive the disease is. Such analyses are only supplementary and can help in choosing the appropriate therapy. AI models are already being trained to take into account both a patient’s facial appearance and the latest results of their tests.
Dr. Raymond Mak of Mass General Brigham, one of the study’s lead authors, says AI models such as FAHR-FaceAge and FAHR-FaceSurvival could help identify people at higher risk of death. Comprehensive testing of those people could make it possible to detect a hidden condition and start treatment earlier, which would change their prognosis.
Zbigniew Wojtasiński (PAP)
