AI Tool Analyzes Selfies to Predict Cancer Survival Risk

Scientists have developed an artificial intelligence (AI) tool, named FaceSurvival, that analyzes facial photographs to predict a cancer patient's risk of mortality. In tests, the tool identified patients at the highest risk of dying within 90 days, with over half of those individuals passing away during that period. In contrast, only two percent of patients classified in the lowest-risk group died within the same timeframe.

The technology, developed by researchers at Mass General Brigham in the US, builds upon earlier findings that cancer patients appearing older than their chronological age were more likely to have a poorer survival outcome. Instead of estimating biological age, FaceSurvival detects subtle facial cues indicative of a patient's prognosis.

According to the study published in The Lancet Digital Health, the tool could potentially transform how physicians assess patients before initiating treatment. Clinicians often make intuitive judgments about a patient's health based on their appearance, a practice known as the 'eyeball test.' Dr. Grace Lee, a study co-author and expert in colon and rectal surgery, stated that the AI tool offers a method to standardize and quantify this intuitive assessment, potentially improving prognostic accuracy and clinical decision-making.

To train the AI, researchers utilized over 40 million publicly available facial photographs. FaceSurvival was then trained to predict mortality using images from more than 34,000 cancer patients. When evaluated on a separate group of 3,820 patients, the tool assigned each a risk score ranging from zero to one. Among patients receiving the highest scores, 64 died within three months, compared to 20 patients in the lowest-risk group. The study found that individuals in the highest-risk quarter were more than three times as likely to die as those in the lowest-risk quarter, even after accounting for age, sex, and cancer type.

These findings were further validated in nearly 4,900 cancer patients in the Netherlands. The research team also developed a second tool, FAHR-FaceAge, which estimates a person's biological age from their facial appearance. Earlier research demonstrated this concept using photographs of actors Paul Rudd and Wilford Brimley, both taken at age 50. Rudd's biological age was estimated at 42.6, while Brimley's, who died at 85, was estimated at 69. In the current study, FAHR-FaceAge was tested on over 38,000 cancer patients. Results indicated that patients whose faces appeared at least five years older than their actual age had a 21 percent higher risk of death, while those who appeared at least five years younger had a 17 percent lower risk.

The study also noted links between lifestyle factors and facial appearance, with smoking, heavy drinking, and drug use associated with an older-looking face. Smoking showed the strongest correlation, with current smokers tending to appear biologically older.

Researchers believe that the combination of FaceSurvival and FAHR-FaceAge could eventually enable faster and more cost-effective patient assessments in clinical settings. However, the technology is still experimental, and further studies are required before widespread hospital implementation. Scientists are also investigating the impact of cosmetic surgery on the AI's accuracy and whether using these tools to guide treatment can ultimately improve patient survival rates.