Study examines AI estimates of cancer survival from facial photographs
Science & ClimateWell-covered
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The Facts
- Mass General Brigham researchers developed AI tools to analyze facial photographs for health-related estimates.
- The research was published in The Lancet Digital Health.
- The FAHR-Face model was trained on more than 40 million facial photographs.
- FAHR-FaceAge estimates biological age from a facial photograph.
- FAHR-FaceSurvival predicts survival outcomes for cancer patients from facial photographs.
- Researchers compared cancer-related estimates with patient data to evaluate accuracy.
- The researchers envision the tools as information that could help guide treatment decisions.
- Experts say the survival tool is not ready for clinical use and needs further testing.
Context
What does biological age mean in this research?
Biological age refers to the body’s health and functional condition rather than the number of years a person has lived. Independent IndexHR 24sata tportal.hr
How was the cancer-survival tool tested?
The model was trained using more than 40 million facial images and photographs of more than 30,000 cancer patients; researchers compared its estimates with patient data. Business Standard Straits Times Independent
Can doctors use this tool now?
No. Reporting on the study says it is not ready for clinical use and should be tested against doctors’ judgment. BBC Straits Times
Where Left and Right agree
Left and right largely agree on this one.
- What both sides accept
- Both framings accept that the survival-prediction tool is unproven for clinical use and that further testing must occur before it touches treatment decisions.
- Where Left and Right differ in emphasis
- The risk of deploying an unvalidated model that predicts life expectancy, or research into cheap, informative diagnostics being needlessly slowed by caution.
- Watch for
- Watch for whether FAHR-FaceSurvival is tested head-to-head against doctors' own clinical judgement, which the Straits Times reports experts say is required before any clinical use.The Straits Times
How left and right read it
Before a model's read of a patient's face is allowed anywhere near a treatment decision, the burden sits with its builders to prove it is valid and fair — not with the patient to object. That burden is not yet met: experts say the survival tool needs further testing before clinical use, even as researchers envision it guiding care. Unproven is unproven. Promise is not permission.
“It is one of the most comprehensive demonstrations to date of how AI analysis of a selfie could help assess a patient's health and potentially guide their medical care” — The Guardian
A photograph costs nothing, so a tool that estimates biological age from one could put real information in the hands of patients and their doctors — which is why work trained on more than 40 million images and checked against patient data deserves to continue. Keep testing it. Experts say the survival model isn't ready for care decisions, so what accuracy would earn it that role?
The receipts — 60 sources
Independent coverage (60)
The news, read by both sides.
The day’s biggest stories in one short brief, built to show both sides and the competing values behind the headlines. Free in your inbox.