The expression ‘the eyes are the mirror of the soul’ is taking on a new, medical meaning. Scientific research shows that the retina contains information not only about vision, but also about the body’s biological ageing. This means it could potentially provide insight into the risk of conditions such as diabetes and cardiovascular disease.
A research team led by Toru Nakazawa of the Tohoku University Graduate School of Medicine has developed an AI model that can estimate so-called ‘retinal age’ based on a single retinal photograph. The results of this study have been published in Communications Medicine and point to a promising role for artificial intelligence in early screening.
AI model predicts age
The AI model was trained on over 50,000 retinal images from healthy adults and validated with a further 7,000 images. Based on visible features in the retina, such as blood vessels and tissue structures, the system calculates the biological age of the eye.
The accuracy is striking: the estimated age deviates by an average of just three years from the actual age. During development, blood sugar levels (HbA1c) were also taken into account to better recognise age-related patterns. In practice, however, no blood test is required; a simple retinal photograph is sufficient.
According to the researchers, this technology can be integrated almost seamlessly into existing healthcare processes. After all, retinal photographs are already routinely taken during eye examinations and general health checks.
‘Retinal age gap’
A key concept in the research is the so-called ‘retinal age gap’: the difference between the retinal age estimated by AI and the actual calendar age. In some patients, this gap proved to be considerably wider.
After adjusting for age and gender, the researchers found that people with conditions such as diabetes, heart disease or a previous stroke were more likely to have an ‘older’ retina than expected. This suggests that accelerated ageing of the retina may be linked to broader health issues.
Although the study primarily demonstrates correlations and does not prove a direct cause-and-effect relationship, this insight offers an important first step towards new screening methods. The retina serves as an accessible biomarker for systemic health in this context.
Accessible screening and prevention
The researchers emphasise that further research is needed to determine the extent to which retinal age can actually predict future disease development. Work is now underway on a large-scale follow-up study in which more than 10,000 participants will be monitored over a three-year period.
The ultimate aim is to deploy this AI application as a screening tool. Healthcare providers would be able to use it to quickly identify which patients would benefit from further investigation or targeted preventive interventions.
In practice, this represents a step towards more proactive and personalised care. By utilising existing imaging and smart algorithms, health risks can be detected earlier, without placing an additional burden on the patient or healthcare provider.
Diagnostics using eye images
There are further examples of clinical presentations or predictions made possible by developments in AI and diagnostics. For instance, last year researchers from Skoltech University and the AIRI Institute developed an AI tool that automates this segmentation, significantly speeding up diagnosis while improving accuracy. The system achieved 97% accuracy and 84% sensitivity, particularly excelling at detecting tiny blood vessels that earlier models often missed. This innovation builds on growing use of AI in eye care, offering faster, more reliable screening and potentially enabling earlier detection of diabetic eye diseases.
A couple of weeks ago we reported on a new system that assesses cardiovascular risk using retinal images captured during routine eye exams. The results show strong alignment with established risk calculators, suggesting that eye care settings could become an additional entry point for preventive cardiology.
