Automated diagnosis of keratitis from low-quality slit-lamp images using an improved generative adversarial…

AI / informaticsNPJ digital medicine· Jul 2026

Automated diagnosis of keratitis from low-quality slit-lamp images using an improved generative adversarial network.

Jiewei Jiang, Siyuan Song, Qihai Yao … Zhongwen Li

Read paper

CorneaAI / informaticsDiagnosisQ1IF 18.0

AI restores low-quality slit-lamp images to accurately diagnose keratitis

Automated diagnosis of keratitis from low-quality slit-lamp images using an improved generative adversarial network.

Jiewei Jiang … Zhongwen Li

NPJ digital medicine · Jul 2026 · Q1 in General Medicine

Background

Keratitis is a major cause of visual impairment worldwide, with timely and accurate diagnosis being critical to preventing irreversible vision loss.

Methods

Automated keratitis diagnosis framework for LQS images demonstrates robust performance when evaluated on a multicenter dataset of 10,498 slit-lamp images, achieving an AUC exceeding 0.949.

n = 10,498 slit-lamp images

Results

AUC exceeding 0.949

restored images let AI tell keratitis from normal eyes with high accuracy

n = 10,498 slit-lamp images

More results

A random degradation algorithm is designed to simulate common clinical artifacts, enabling paired LQS-HQS training.

“

Conclusion

This approach presents a practical solution for keratitis diagnosis from degraded images, with strong potential to reduce diagnostic disparities in resource-limited environments.

Automated diagnosis of keratitis from low-quality slit-lamp images using an improved generative adversarial network.

Automated diagnosis of keratitis from low-quality slit-lamp images using an improved generative adversarial network.

Related papers