Study: Deep learning models show promise in diagnosing infectious keratitis

(Image credit: Adobe Stock/K_Ho)

A new study suggests that artificial intelligence (AI) may support eye care specialists in diagnosing infectious keratitis (IK)—a leading cause of corneal blindness worldwide. According to the findings, deep learning (DL) models demonstrated similar levels of diagnostic accuracy to ophthalmologists.

In a meta-analysis published in eClinicalMedicine, Darren Ting, PhD, FRCO, MBChB, PGCHPE, from the University of Birmingham, led a team of global researchers in reviewing 35 studies that used DL models to diagnose infectious keratitis. The AI models exhibited a sensitivity of 89.2% and a specificity of 93.2%, compared with ophthalmologists' sensitivity of 82.2% and specificity of 89.6%.¹

Ting, is senior author of the study, a Birmingham Health Partners (BHP) Fellow and consultant ophthalmologist at the University of Birmingham

The research team noted the DL models analyzed more than 136,000 corneal images, underscoring AI's potential as a diagnostic tool in clinical settings.

“Our study shows that AI has the potential to provide fast, reliable diagnoses, which could revolutionize how we manage corneal infections globally," Ting said in a news release. "This is particularly promising for regions with limited access to specialist eye care and can help reduce the burden of preventable blindness worldwide."

The study also found that AI models were effective at distinguishing between healthy eyes, infected corneas, and the various causes of IK, including bacterial and fungal infections. However, the authors stressed the need for more diverse data and further external validation to ensure these models’ reliability in clinical practice.¹

According to the researchers, infectious keratitis, an inflammation of the cornea, impacts millions, particularly in low- and middle-income countries where specialist eye care is scarce. As AI technology evolves and plays a larger role in medicine, it may soon become an essential tool in combating corneal blindness globally.

Reference:

1. Ong, Zun Zheng et al. Diagnostic performance of deep learning for infectious keratitis: a systematic review and meta-analysis. eClinicalMedicine, Volume 77, 102887. Accessed October 28, 2024. DOI: 10.1016/j/eclinm.2024.102887.