Diabetic retinopathy (DR) is a rapidly rising cause of blindness, putting hundreds of millions of people who are struggling with diabetes at risk of losing their eyesight.
While irreversible at a late stage, DR is treatable if caught early enough. To that end, doctors are getting help catching this disease from an unlikely source: an AI.
Researchers at Google have developed a method for artificial intelligence to analyze and detect DR in patients, potentially allowing a greater number of patients to be tested for the disease – especially in areas without greater access to professional healthcare.
After analyzing approximately 128,000 retinal images, the findings showed that Google's machine learning algorithm exhibited sensitivity and specificity "on-par with that of ophthalmologists" when looking for minuscule lesions and hemorrhages in the retina.
While this is a big step for both tech and medicine, there's still work to do for Google's machine learning team. The group hopes to continue working with experts to improve the algorithm's performance, as well as finding ways to work with screening staff to efficiently use the technology to treat patients.