Since antiquity, the eyes have been considered a 'window to the soul', through a symbolic transfiguration of a philosophical nature, but today, thanks to artificial intelligence (AI), we will soon be able to consider them a window to the health of the entire organism. This type of diagnostic paradigm has already been supported in the past by clinical evidence emphasising the connection between the retina, our photosensitive film, and various organs, including the heart, brain, kidneys, lungs, and also with processes such as ageing or metabolic changes.
These interconnections were the subject of very recent research published in Nature Communications, "Photoreceptor metabolic window unveils eye-body interactions" with which a new model based on artificial intelligence was presented, which can be used to link retinal conditions to a range of multi-system diseases, including cardiovascular, metabolic, renal and pulmonary disorders.
To provide a practical tool, the "Photoreceptor Metabolic Window (PMW)an artificial intelligence-based algorithm, with which it was possible to examine prospective associations between retinal photoreceptor layer thickness, a range of metabolites and 20 systemic pathological conditions.
An extensive cohort of patients from different countries was recruited, in particular, data from more than 120,000 participants, acquired through the UK Biobank, were examined and the results were validated by cross-referencing them with the GDES (Guangzhou Diabetic Eye Study).
One must, first of all, consider that while the barrier between the blood circulation and the brain is very filtering, the barrier between the retina and the blood circulation can be considered permeable to many substances, including certain lipoproteins from the bloodstream. This implies that the retina is not isolated and metabolic disturbances, which have an overall impact on the various organs, due to the exceptional susceptibility of the photoreceptors - which are among the most metabolically active cells - could be the site of the first manifestations of damage from minor metabolic stresses.
109 metabolites, circulating in the blood, were identified and found to be correlated with the health status of retinal photoreceptors. Thinning of the photoreceptor layer was also found to be associated with an increased risk of 13 major diseases, such as lung cancer, chronic obstructive pulmonary disease, diabetes mellitus and terminal nephropathy.
Through metabolic profiling, the researchers found that certain amino acids and lipids were crucial for photoreceptor integrity. Low levels of omega-3 fatty acids and phospholipids correlated with thinning of the photoreceptor layer and increased risk, while high creatinine levels were associated with an overall risk of deteriorating systemic health
Based on these pioneering results, the researchers propose as a potential screening tool for systemic diseases a systematic use of retinal imaging, which has the advantages of non-invasiveness and low cost. When an abnormality is identified early during a routine eye examination, clinicians could implement preventive interventions in patients at higher risk of developing systemic diseases.
It must be emphasised that we are only at the beginning of this promising strand of bio-diagnostics and that this type of screening will only become fully operational after further research and clinical studies. In any case, it must be emphasised that AI-based approaches are bringing about a real revolution in preventive medicine and may in the future lead to a stage where accurate retinal examination has the potential to be truly considered a key indicator of a patient's overall health condition.
On the subject of AI applications, we point out
- Generative AI in Ophthalmology - Oculista Italiano
- AI and ophthalmology: listen to the podcast - Oculista Italiano
- Geographical atrophy & AI - Oculist Italiano
- AI & Oculomics - Oculista Italiano
- Yang, S., Xin, Z., Cheng, W. et al.Photoreceptor metabolic window unveils eye-body interactions. Nat Commun 16, 697 (2025). https://doi.org/10.1038/s41467-024-55035-x
- Zhou, Y. et al. A foundation model for generalizable disease detection from retinal images. Nature622, 156-163 (2023).
English 
Italian