Artificial intelligence is showing new potential as a tool for preventive medicine. Researchers report that a deep learning model can estimate how fast the body is ageing by analysing routine chest X-rays, often more accurately than widely used DNA-based ageing measures.
In the study, researchers evaluated a deep learning algorithm known as CXR-Age, which extracts age-related signals from standard chest X-ray images. Its performance was compared with two established epigenetic “ageing clocks” based on DNA methylation: Horvath Age and DNAm PhenoAge. The analysis included data from 2,097 adults enrolled in the Project Baseline Health Study, a large, multi-centre US research programme focused on long-term health monitoring.
Imaging-based ageing signals
The AI model demonstrated strong associations with early, subclinical signs of ageing in the heart and lungs. Higher CXR-Age scores correlated with the presence of coronary artery calcium, reduced lung function, increased frailty and elevated blood proteins linked to inflammation, neurodegeneration and biological ageing. Notably, these associations were already visible in middle-aged adults.
By contrast, the DNA-based epigenetic clocks showed weaker or inconsistent relationships with these early indicators of cardiopulmonary decline, particularly in individuals who had not yet developed overt disease.
According to the researchers, this suggests that medical imaging may capture functional and structural changes in organs that are not fully reflected in molecular ageing markers alone. Their findings were published in The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences.
Towards earlier risk identification
“Deep learning applied to common clinical images can reveal how organs are ageing beneath the surface,” said Douglas P. Kiel, MD, MPH, co-author of the study and director of the Musculoskeletal Research Center at the Marcus Institute for Aging Research. “This type of information could eventually help clinicians identify people at risk of age-related disease before symptoms appear.”
Because chest X-rays are already widely used in clinical practice, an AI-based ageing metric could be integrated into existing workflows with relatively low barriers. Rather than replacing current risk assessments, the researchers see CXR-Age as a complementary tool that adds a new layer of insight into cardiopulmonary health and biological ageing.
Implications for personalised care
The study highlights the growing role of AI and medical imaging in personalised and preventive healthcare. By translating subtle image features into meaningful biomarkers, AI systems like CXR-Age may support earlier intervention, more precise risk stratification and better monitoring of ageing-related health trajectories.
While further validation and clinical studies are needed, the findings underscore how routinely collected imaging data, when combined with advanced machine learning—could help shift healthcare from reactive treatment towards proactive prevention.
