AI Facial Photo Models for Cancer Survival and Biological Age

This news segment examines research from Mass General Brigham on AI models that analyze facial photographs to estimate cancer survival risk and biological age. NBC News reporters discuss the study design, the facial regions highlighted by the models, and the medical and ethical questions surrounding the work.

■ AI cancer-survival model
- Trained on tens of millions of facial images and evaluated on about 38,000 cancer patient photos
- Regions below the eyes and the nasal bridge were linked to mortality risk
- High-risk and low-risk groups showed differences in 90-day survival

■ FaceAge biological-age model
- Trained on about 750,000 images
- Highlighted regions included nasolabial folds, forehead, and temples

■ Medical and clinical considerations
- What the AI detects in these facial areas remains unclear
- The models are not ready for clinical practice; patient communication and treatment decisions require more research

■ Bias and dataset limitations
- Data came from Harvard and Netherlands datasets; the test set included far more white patients than Black patients

Viewers interested in AI in healthcare, oncology prognostics, and algorithmic bias can gain an overview of the research claims and open questions. The segment does not provide clinical guidance or validated tools, so the findings should be treated as early research.