Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Chest X-rays Reveal Hidden Aging: Deep Learning Model Predicts Mortality Risk
A deep learning model reveals biological aging patterns on routine chest X-rays, identifying individuals at significantly elevated risk for premature death—with diagnostic power exceeding traditional risk factors.
Background
Chronological age is a crude measure of health risk; individuals age biologically at vastly different rates. Researchers developed AgeNet, a deep learning model trained on healthy individuals, to estimate “radiographic age” from chest radiographs. This study evaluated whether this non-invasive biomarker predicts mortality in 421,894 Korean adults aged 40-80.
Key Findings
Two aging metrics independently predicted mortality:
- Accelerated aging (radiographic age ≥5 years above chronological age): increased all-cause mortality 26% in males and 52% in females (P<.001)
- Aging velocity (annual change in radiographic age): accelerated aging velocity (≥1.5 years/year) increased mortality risk 51-71% independent of baseline status (P<.001)
- Decelerated aging was protective, reducing mortality 22-24%
- Strongest associations observed with cardiovascular and respiratory mortality
Why It Matters
Chest radiographs are inexpensive, non-invasive, and already routine in clinical practice. Converting them into objective biological aging biomarkers enables cost-effective mortality risk stratification for targeted interventions—potentially transforming how we identify individuals needing preventive care.
Limitations
Findings require validation in non-Korean populations. Biological mechanisms linking radiographic aging patterns to cause-specific mortality remain unexplored.
Original paper: Accelerated Aging and Aging Velocity from Deep Learning-based Chest Radiograph-derived Age for Predicting Cause-specific Mortality. — Radiology. Artificial intelligence. 10.1148/ryai.250609
