One of the most impressive areas of generative AI software like ChatGPT right now involves enhanced computer vision. AI can understand and interpret data from images. That’s why we now have such advanced image and video generation models in ChatGPT, Gemini, Firefly, and other AI software.
Models like ChatGPT o3 can accurately guess the location of an image by analyzing its details. Google offers advanced photo editing tools in its Photos app, and also directly in Gemini. These tools let you alter real photos in ways that weren’t possible before.
These image-related AI capabilities aren’t just used to generate memes or melt OpenAI’s servers. Researchers are developing AI models that can interpret images for various purposes, including medicine.
The latest study showing such advancements comes from China.
Researchers from several universities have been able to determine a person’s age with high accuracy by having AI inspect an image of their retina. The readings also showed differences between the person’s age and the eye’s age. The researchers found that the retinal age gap the AI provided can be especially helpful for women. A simple retinal scan might help doctors offer better support to couples trying to conceive and to women at risk of early menopause.
Retinal fundus imaging, or a photo of the back of the eye, lets doctors see microvascular features that reflect systemic aging. An AI trained on thousands of images can then predict the eye’s age and compare it to the person’s actual age to “predict retinal age from fundus images with high precision.”
The scientists used an AI called Frozen and Learning Ensemble Crossover (FLEX) to predict retinal age from fundus images. They fed FLEX over 20,000 eye photos from more than 10,000 adults of all ages to teach it how the back of the eye looks as people age. FLEX also analyzed over 2,500 images from nearly 1,300 pre-menopausal women.
The AI was then able to estimate a person’s age by examining a retinal fundus photo. If the eye appears older than the woman’s actual age, the retinal age gap is positive. That could also mean other organs in the body are aging faster.
The implications for reproductive health are clear. Fertility and menopause issues could benefit directly from such an AI screening tool.
The researchers linked a larger retinal age gap to lower blood levels of anti-Müllerian hormone (AMH), a marker for ovarian reserve. The lower the AMH value, the harder it is for older women to conceive.
The scientists studied women ages 40 to 50 and found that each additional retinal year raised the risk of a low AMH result. The risk increased by 12% in the 40-44 age group and by 20% in the 45-50 group for every extra retinal year.
The study also found that having more childbirths at younger ages was associated with lower AMH levels than average.
Each additional retinal year increased the risk of developing menopause before age 45 by 36%, according to the paper.
We’re still in the early days of using AI for medical imaging, but the study shows promise for using a simple, non-invasive technique to improve reproductive health protocols. Imagine getting a retinal scan in your late 20s or early 30s to help decide whether to get pregnant or freeze your eggs.
Similarly, women over 40 concerned about pre-menopause or menopause could use an eye scan to check their retinal age and assess the risk of early symptoms. This might help them prepare for the years ahead with hormonal therapies to delay or ease symptoms.
For any of this to happen, the conclusions from Hanpei Miao & Co. would need to be confirmed by further research.
Separately, the FLEX AI model used in this study could be explored for other health conditions where eye scans might serve as early indicators of age-related health risks.
The full study is available in Nature magazine.
