Google’s New AI Research Uses Deep Learning on Retinal Pictures to Create an Age Predictor

The physiologic and molecular changes associated with becoming older raise a person’s odds of being sick and dying. Researchers can find ways to lessen the prevalence and severity of diseases by measuring and estimating the biological markers of aging. To distinguish between a person’s biological age and their chronological age, scientists have devised “aging clocks” that use biomarkers like blood proteins or DNA methylation to estimate a person’s biological age. These aging clocks can estimate the risk of developing an age-related illness. However, due to the need for a blood sample, other methods of locating equivalent measurements may make aging data more accessible.

Recent research published under “Longitudinal fundus imaging and its genome-wide association analysis provide evidence for a human retinal aging clock” demonstrates that deep learning models can reliably estimate a person’s biological age from a retinal image and provide new insights into the prediction of age-related diseases. Researchers are also making available the updated source code for these models based on previously disclosed ML frameworks for processing retina pictures.

Age estimation using retinal photographs

Multiple primary care clinics’ worth of de-identified retinal pictures were used to train a model to predict chronological age for participants in a telemedicine-based blindness prevention program. The resultant model’s performance was measured against a hidden dataset of 50,000 retinal pictures and the UKBiobank’s main dataset of around 120,000 images. Given the moniker eyeAge, the model’s projections agree quite well with people’s chronological ages.

An accurate aging clock based on retinal pictures has never been made before.

Comparison between Expected and Actual Age Difference

Although eyeAge has a high correlation with chronological age across many samples, there are still certain situations when the model predicts a number significantly younger or older than the chronological age, as seen in the figure above. This may suggest that the model is picking up on details in the retinal pictures indicative of real-world biological consequences pertinent to age-related disorders.


  • This facilitates the identification of genes whose activities might be altered by medications to promote healthy aging and the discovery of indicators for aging and age-related disorders.
  • The impact of lifestyle behaviors and treatments like exercise, food, and medicine on biological aging may also be better understood.
  • The eyeAge clock might be used in the pharmaceutical sector to gauge the efficacy of rejuvenation and anti-aging drugs.
  • Researchers may ascertain whether or not these therapies successfully delay or reverse the aging process by monitoring changes in the retina over time.

The results also show that the blood-biomarker-based aging clock cannot be compared to the retina-based aging clock used by eyeAge. EyeAge might be utilized for actionable biological and behavioral therapies in contrast to conventional aging clocks since imaging is less intrusive than blood testing. When paired with other indicators, it gives a complete knowledge of an individual’s biological age and allows researchers to examine aging from a new perspective.

Predictive aging clocks have been used to learn more about biological age, which differs from a person’s chronological age. However, their precision in shorter periods could be much better. In this study, researchers used fundus photos from the EyePACS dataset to train deep-learning models to estimate people’s ages. Compared to alternative aging clocks, ‘eyeAge”s retinal aging clocking predicted chronological age more accurately (mean absolute error of 2.86 and 3.30 years on quality-filtered data from EyePACS and UK Biobank, respectively).

When controlling for phenotypic age, the hazard ratio for all-cause mortality in eyeAge remained at 1.026, demonstrating its independence from blood marker-based assessments of biological age. Multiple GWAS findings in the UK Biobank sample supported eyeAge’s uniqueness among individuals. Alk knockdown in flies helped the top GWAS locus by reversing the age-related deterioration in fly eyesight. This research illustrates the promise of a retinal aging clock as a tool for researching aging and age-related disorders and quantitatively quantifying aging on extremely short time scales, paving the way for rapid and actionable evaluation of gero-protective medicines.

Check out the Paper and Google blog. All Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to join our 18k+ ML SubRedditDiscord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.

🚀 Check Out 100’s AI Tools in AI Tools Club

Dhanshree Shenwai is a Computer Science Engineer and has a good experience in FinTech companies covering Financial, Cards & Payments and Banking domain with keen interest in applications of AI. She is enthusiastic about exploring new technologies and advancements in today’s evolving world making everyone's life easy.

🚀 LLMWare Launches SLIMs: Small Specialized Function-Calling Models for Multi-Step Automation [Check out all the models]