A meta-analysis by an international team of researchers found artificial intelligence (AI)-driven deep-learning (DL) models achieved higher accuracy than ophthalmologists in diagnosing infectious keratitis from slit-lamp images. 

Senior author Honorary Associate Professor Darren Ting, University of Birmingham, UK, and colleagues analysed 35 studies (136,401 corneal images from >56,011 patients). DL showed a sensitivity of 89.2% and specificity of 93.2%, versus ophthalmologists’ 82.2% and 89.6%, respectively. Based on anterior segment photography, the DL models also achieved a sensitivity of 96.9% and a specificity of 96.7% in diagnosing/distinguishing keratitis from healthy corneas and other pathologies, and differentiated between bacterial, fungal, acanthamoeba and viral keratitis with 86.2% sensitivity and 83.6% specificity. 

The study shows AI has the potential to provide fast, reliable diagnoses, which could revolutionise how we manage corneal infections, said A/Prof Ting. "This is particularly promising for regions with limited access to specialist eyecare and can help reduce the burden of preventable blindness worldwide."