Newly approved software uses AI to detect diabetic retinopathy

The FDA has approved marketing of new software, called IDx-DR that utilises artificial intelligence to detect at least moderate diabetic retinopathy.

Clinicians first take digital images of the eye with the Topcon NW400 retinal camera and then upload the images to a cloud server with IDx-DR. Using an AI algorithm, the software determines whether "more than mild diabetic retinopathy" is detected or not. If it is detected, patients are referred to an eye care specialist for further evaluation and treatment.

The FDA says the system may be useful for clinicians not normally involved in eye care since they do not need to interpret the images.

In a study of 900 patients with diabetes, the software had 87% accuracy in identifying patients with more than mild diabetic retinopathy and 90% accuracy in classifying those who did not.