AI uses selfies to predict cancer survival, study shows

Researchers have developed an algorithm that uses a person's photo to predict biological age and survival in cancer patients. The results of the technology, called FaceAge, were published Thursday (8) in the journal The Lancet Digital Health.

From the researchers' point of view, the work demonstrates that a photo, such as a simple selfie, contains important information that can help in clinical decision-making and treatment plans for patients and doctors.

Biological age is that which reflects the functional condition of the human body, and is not necessarily aligned with a person's chronological age. For example, a sedentary individual who smokes and drinks alcohol may have a more advanced biological age, due to an unhealthy lifestyle, than their age in relation to their date of birth.

According to the study, FaceAge predictions indicating older biological age were associated with worse overall survival outcomes across multiple cancer types. They also found that the algorithm outperformed doctors in predicting short-term life expectancies for patients who received palliative radiation therapy.

“We can use artificial intelligence (AI) to estimate a person’s biological age from facial photos, and our study shows that the information can be clinically meaningful,” study co-senior and corresponding author Hugo Aerts, director of the Artificial Intelligence in Medicine (AIM) program at Mass General Brigham, said in a statement.

How was AI developed?

Researchers at Mass General Brigham, a US hospital research company, used deep learning and facial recognition technologies to train FaceAge. The tool was trained on 58,851 photos of supposedly healthy individuals from public datasets.

The team tested the algorithm on a cohort of 6,196 cancer patients from two centers, using photographs taken routinely at the start of radiotherapy treatment.

The results showed that cancer patients appeared significantly older than those without cancer, and their FaceAge, on average, was about five years older than their chronological age.

Among cancer patients, more advanced FaceAge was associated with worse survival outcomes, especially in individuals who appeared older than 85 years, even after adjustments for chronological age, sex, and cancer type.

Next steps

More research is needed before this technology can be considered for use in a real clinical setting. The team is currently testing the algorithm to predict disease, overall health status and life expectancy.

Follow-up studies include expanding this work to different hospitals, analyzing patients at different stages of cancer, tracking FaceAge's estimates over time, and testing its accuracy on plastic surgery and makeup data sets.

“This opens the door to a new field of biomarker discovery from photographs, and its potential goes far beyond cancer treatment or age prediction,” said study co-senior author Ray Mak, a faculty member in the AIM program at Mass General Brigham, in a statement.

“As we increasingly consider many chronic diseases as diseases of aging, it becomes even more important to be able to accurately predict an individual’s aging trajectory. I hope that we can ultimately use this technology as an early detection system in a variety of applications, within a sound regulatory and ethical framework, to help save lives.”

AI in medicine: early diagnosis and risk reduction for patients