Deep-learning algorithm detects diabetic retinopathy on par with regional specialists

New research out of Thailand found that a deep learning algorithm was able to diabetic […]

New research out of Thailand found that a deep learning algorithm was able to diabetic retinopathy in patients with diabetes on par with community specialists, according to a study published in The Lancet.

Researchers discovered that the deep-learning system was able to detect vision-threatening diabetic retinopathy with an accuracy of 94.7%, sensitivity of 91.4% and specificity of 95.4%. This was compared to the retina specialist over-readers who performed with an accuracy of 93.5%, a sensitivity of 84.8% and a specificity of 95.5%.

“A deep-learning system can deliver real-time diabetic retinopathy detection capability similar to retina specialists in community-based screening settings,” researchers wrote. “Socioenvironmental factors and workflows must be taken into consideration when implementing a deep-learning system within a large-scale screening programme in [low-income and middle-income countries].”

In order to be included, patients had to be listed on the national diabetes registry of Thailand and at least 18. Patients also needed to be able to have their “fundus photograph taken for at least one eye and due for screening” as part of the country’s national guidelines. The research included 7,651 patients.

Every image was read by the machine learning algorithm, as well as a regional retinal specialist. Additionally, images sampled for adjudication ​​reference standard were also reviewed by panels of three U.S. board-certified specialists.

The study, which was co-developed and co-funded by Google, took place in central Bangkok, Chiang Mai, and Pathum Thani, Thailand.


Individuals with Type 1, Type 2 or gestational diabetes can develop diabetic retinopathy, which can cause vision problems or even blindness, according to the CDC.

The CDC reports that diabetic retinopathy is the leading cause of blindness in adults in the U.S. The condition impacts roughly 4.1 million in the U.S. Treatment for the condition may include medication injections, laser treatment and eye surgery, according to the National Eye Institute.


Increasingly technology has been put to work detecting diabetic retinopathy. In 2019, Alphabet’s life science subsidiary Verily unveiled real-world clinical use for its machine learning algorithm to screen for diabetes-related conditions including diabetic retinopathy and diabetic macular edema. The tool was first rolled out in India and aimed at helping with the shortage of doctors worldwide.

There is a history of research on AI for diabetic retinopathy screening. For example, in 2020 The Lancet Digital Health published a study that found automated or semi-automated deep learning systems for diabetic retinopathy screening could lead to cost savings at a health-system level.

Original Article: (