Revolutionising Diabetic Eye Care: Harnessing the Power of AI

AI in DR

AI has massive potential in the field of ophthalmology, given the heavy reliance on imaging for diagnosing ocular disease, and companies have been working with AI for over 20 years to try and reduce the reliance on human graders.

Deep learning techniques are fed into AI models that learn from a large dataset of retinal images, and this capability empowers the program to make predictions.

The US Food and Drug Administration (FDA) has already cleared three AI algorithms for use in DR screening: IDx-DR, EyeArt, and AEYE Diagnostic Screening. In the EU, there are more, including Google and Singapore Eye Lesion Analyzer (SELENA).

These AI systems are fully autonomous algorithms that work without human supervision and, in trials, have impressive sensitivity and specificity performances (Table 1).

Several AI devices have the potential to influence DR screening considerably, and some have demonstrated promising performance on prospective data sets. However, before achieving optimal care for individuals with diabetes, it is imperative to address existing knowledge gaps.

Additional head-to-head validation studies are urgently needed to provide clinicians with essential insights to determine the most effective AI devices for integration into their practice.

Not so fast

AI is an incredibly powerful tool still in the early stages; policies to ensure that the use of AI in healthcare aligns with recognised standards of safety, efficacy, and equity is a must-have.

The authors acknowledge the challenges in deploying AI for DR screening, including data acquisition, bias in data, difficulty in comparing different algorithms, and human barriers to AI adoption in healthcare.

It is also important to note the risks of AI biases, such as high false positive rates, underdiagnosis, and the patient’s anxiety or belief in their AI-supported diagnosis.

Why does it matter?

The global prevalence of DR in 2020 was approximately 103 million individuals. With a projected rise to 161 million by 2045, the global DR burden will likely remain consistently high, with countries in the Middle East, North Africa and the Western Pacific being disproportionately affected.

Treatment and early intervention for DR are crucial because they could help prevent or slow down the progression, preserving vision.

Vision loss can profoundly impact an individual’s quality of life, affecting their ability to perform daily tasks, work, and engage in activities they enjoy. Treating DR early due to effective diagnosis, aided by AI, can help maintain or improve the quality of life for individuals.

Take Home Messages

1. AI can leverage deep learning techniques for diabetic retinopathy.
2. AI use in DR could significantly increase the rate of early diagnosis while being cost-effective.
3. With caveats, AI has a transformative potential in diabetic retinopathy.