🔍 AI Can Predict Nearsightedness from a Single Eye Scan

Researchers in China trained a neural network to forecast the progression of myopia years in advance. The model was trained on retinal images from more than 3,000 schoolchildren who underwent annual eye exams over five years.

The system can determine:

✅ Whether a child's vision will worsen with 86% accuracy.
✅ Whether high myopia (−6.0 diopters or worse) will develop with 98% accuracy.
✅ What a person's prescription is likely to be in one, two, three, four, and five years.

In long-term forecasts, accuracy depends on the time horizon. A one-year prediction has an average error of just ±0.37 diopters, while a five-year forecast comes with an error of about ±1.1 diopters. However, if the AI is given images collected over multiple years instead of a single snapshot, the error drops to ±0.32 diopters—roughly the equivalent of one step in eyeglass prescription strength.

💡 Why This Matters?

Myopia is becoming more common among teenagers, and rapid vision decline can lead to serious complications such as glaucoma, retinal detachment, and even blindness. Predicting changes years in advance helps doctors identify which patients need closer monitoring to prevent long-term damage.

@hiaimediaen