Mass General Brigham AI Reads Mortality Risk in Facial Photographs
On October 19, researchers from Mass General Brigham published findings in The Lancet Digital Health regarding two AI systems utilizing FAHR-Face, which were trained on more than 40 million facial images to assess biological age and predict cancer patient survival. FAHR-FaceAge evaluates biological age, while FAHR-FaceSurvival focuses on the nose bridge and under-eye area to estimate mortality risk. The latter identified 122 cancer patients at high risk, with 52.5% succumbing within 90 days, in stark contrast to around 2% in the low-risk category. Additionally, FaceAge revealed that 21% of 1,060 patients estimated to be over 20 years older died within 90 days, compared to 6% of 4,195 who were assessed to be at least 10 years younger. Participants were primarily from the U.S. and Netherlands, mostly aged in their 50s and 60s.
Key facts
- Findings published October 19 in The Lancet Digital Health by Mass General Brigham researchers in Boston
- FAHR-Face model trained on more than 40 million facial images across ages and ethnic backgrounds
- FAHR-FaceSurvival focuses on the bridge of the nose and beneath the eyes
- 52.5% of 122 high-risk cancer patients died within 90 days, versus about 2% of 975 lowest-risk patients
- 21% of 1,060 patients estimated over 20 years biologically older died within 90 days, versus about 6% of 4,195 estimated at least 10 years younger
- Patients came from the United States and the Netherlands, from children to adults in their 90s, mostly in their 50s and 60s
- Heavy alcohol use, drug use and smoking were associated with appearing biologically older
- Rhinoplasty produced a small change in estimates; lighting, expression, pose and makeup had little effect
Entities
Artists
- Raymond Mak
Institutions
- Mass General Brigham
- Mass General Brigham Cancer Institute
- The Lancet Digital Health
- Artificial Intelligence in Medicine program
- CNN
- BBC
- PetaPixel
Locations
- Boston
- United States
- Netherlands