Summary: Researchers have introduced an AI tool named FaceAge, designed to assess biological age from facial images and predict survival rates in cancer patients. A study with over 6,000 participants revealed that cancer patients exhibited FaceAges roughly five years older than their actual age, with higher FaceAges correlating to lower survival rates.
The tool surpassed traditional clinical methods in estimating short-term life expectancy for patients undergoing palliative radiotherapy, especially when its insights were included in decision-making. These results indicate that facial characteristics could act as significant, non-invasive biomarkers for aging and disease, potentially revolutionizing precision medicine.
Key Facts:
- FaceAge AI: Estimates biological age and survival likelihood using facial photographs.
- Cancer Insight: Cancer patients seemed roughly five years older than their chronological age.
- Clinical Boost: FaceAge enhanced doctors’ ability to predict life expectancy in palliative settings.
Investigators at Mass General Brigham developed FaceAge, leveraging deep learning to analyze facial images to gauge biological age and survival chances for cancer patients.
The study revealed that cancer patients had higher FaceAges, implying they looked around five years older than their chronological age.
Additionally, higher FaceAge predictions were linked to poorer survival rates across various cancer types.
The tool outperformed medical professionals in predicting short-term life expectancies for those receiving palliative treatment.
Co-senior author Hugo Aerts, PhD, highlighted the significance of the study, pointing out that leveraging AI in this manner can yield clinically valuable insights.
Aerts noted, “The appearance of a person’s face can provide critical information for shaping treatment decisions. Patients who look younger than their chronological age tend to fare better following cancer treatments.”
While aspects of appearance may offer physicians valuable insights into patients’ overall health, biases regarding age can affect assessments, underscoring the necessity for objective measures.
The researchers trained FaceAge using 58,851 photos from presumed healthy individuals sourced from public datasets, and then tested it on 6,196 cancer patients’ images taken at the beginning of their radiotherapy.
Results indicated that cancer patients typically displayed a significantly older appearance than non-cancer patients, with the average FaceAge being about five years older.
Among cancer patients, increased FaceAge was associated with worse survival, particularly in those believed to be over 85, even when controlling for chronological age and cancer diagnosis.
Estimating survival time in cancer care can be complex, yet it’s crucial for treatment planning. Clinicians attempted to predict the life expectancy of 100 patients undergoing palliative treatment based on their photographs.
Although clinician performance varied, their predictions were only marginally better than random guesses, even with context like chronological age and cancer details. Interestingly, when they were informed of a patient’s FaceAge, their predictions improved significantly.
Future investigations will further explore this technology’s potential in clinical settings by assessing its accuracy across various diseases and stages of cancer.
“This research opens pathways for discovering biomarkers from photographs, yielding applications beyond cancer care,” remarked co-senior author Ray Mak, MD. “As chronic diseases become increasingly viewed through the lens of aging, accurately forecasting an individual’s aging path is becoming more crucial.”
Funding: The research was funded by the National Institutes of Health and the European Union’s European Research Council.
