Can AI help prevent diabetes-related vision loss?

Diabetes is a condition that can have serious health consequences, including the risk of vision loss. But can artificial intelligence (AI) play a role in preventing this complication? In this article, we explore the possibilities and challenges of AI in the fight against diabetic retinopathy.

Terry Quinn's Story

Terry Quinn was diagnosed with diabetes as a teenager. While he was concerned about possible complications, losing his sight was not at the top of his list of concerns. Nevertheless, this eventually became a reality for him.

“I never thought I would lose my sight,” says Quinn. But when he noticed bleeding in his eye, he was diagnosed with diabetic retinopathy. Despite treatments, his vision deteriorated significantly, which had a major impact on his daily life.

The importance of early screening

In the United Kingdom, the NHS invites patients for diabetic eye exams every one to two years. In the United States, it is recommended that adults with type 2 diabetes be screened annually. Dr. Roocasa Channa, retinal specialist at the University of Wisconsin-Madison, emphasizes, “There is very clear evidence that screening prevents vision loss.”

However, many patients do not receive the recommended screenings in practice. Barriers such as costs, communication and accessibility play a role here.

The role of AI in screening

This is where AI comes into the picture. By using artificial intelligence, the screening process for diabetic retinopathy could become more efficient and less expensive. AI systems can be trained to analyse fundus images (photos of the back of the eye) and identify potential problems.

Companies such as Retmarker and Eyenuk have developed AI systems that show promising results. João Diogo Ramos, CEO of Retmarker, explains: “Normally, we use it more as a supportive tool to provide information to people to make a decision.”

Challenges and Limitations

While AI systems are promising, there are also challenges. Google Health research showed that their AI system performed differently in Thailand than in hypothetical scenarios. Problems such as dirty lenses, unpredictable lighting, and varying camera operators' expertise affected the results.

In addition, AI systems can sometimes give false positive results, leading to unnecessary anxiety and costs. It is therefore important to find the right balance between sensitivity (detecting disease) and specificity (correctly identifying healthy eyes).

Cost-effectiveness and accessibility

Cost-effectiveness is a crucial factor in implementing AI systems. In Singapore, a hybrid model, where AI performs initial screening and people verify results, was found to be the most cost-effective.

However, Dr. Bilal Mateen, chief AI officer at PATH, stresses that cost-effectiveness can vary widely between countries. “We need more than just effectiveness data for effective decision making,” he says.

Future perspective

Despite the challenges, Dr. Channa remains optimistic about the future of AI in diabetic retinopathy screening. “I would like to see all our patients with diabetes screened in a timely manner. And I think that given the burden of diabetes, this is a really potentially great solution.”

It's important to remember that AI can't detect all eye problems. Regular check-ups with an ophthalmologist remain essential, especially for older people and people with existing eye problems.

Conclusion

AI offers promising opportunities to help prevent diabetes-related vision loss. By making screenings more accessible and efficient, AI can be a valuable addition to current care. However, there are still challenges to overcome, such as improving accuracy and ensuring cost-effectiveness and accessibility for all. With ongoing research and development, AI can play a critical role in preserving the vision of millions of people with diabetes worldwide in the future.