Chinese researchers created a new artificial intelligence platform that not only has a high success rate in diagnosing cataracts but can also recommend which cases should be referred for further investigation.

The AI system’s training and validation datasets included 37,638 slit lamp photos from an ongoing national Chinese cataract screening program by the Chinese Medical Alliance for Artificial Intelligence.

The study labeled the datasets using a three-step strategy. First, it identified the image capture mode (mydriatic-diffuse, mydriatic-slit lamp, non-mydriatic-diffuse and non-mydriatic-slit lamp). Next, the AI made a clinical assessment (normal, cataract or postoperative eye). Finally, it detected referable cataracts based on etiology and severity.

Investigators said they integrated their cataract AI system with a real-world referral pattern that included self-monitoring at home, or referral to primary healthcare or specialized hospital services.

Researchers reported high success rates in all three of the AI system’s tasks:

• Capture mode recognition: 99.28% to 99.71%.
• Cataract diagnosis: 99.82% for normal lenses, 99.96% for cataract and 99.93% for postoperative eye for mydriatic-slit lamp mode and more than 99% for other capture modes.
• Detection of referable cataracts: Over 91% in all tests.

The AI system recommended 30.3% of people should be referred, which represented a substantial increase in the ophthalmologist-to-population service ratio 10:2 fold compared with the traditional pattern, the study noted.

In contrast to systemic diseases or other ocular disorders, cataracts hold promise for being managed by AI given their uniform lesion areas and pathological base (i.e., cloudy lens), investigators said.

The research team suggested their system could be extended to the management of other ophthalmic diseases that would include updated user-accessible mobile devices and automatic examination instruments.

“The universal AI platform and multilevel collaborative pattern showed robust diagnostic performance and effective service for cataracts,” the researchers wrote in their paper. “The context of our AI-based medical referral pattern will be extended to other common disease conditions and resource-intensive situations.”