A team of researchers from the US has examined the performance of an automated deep learning algorithm compared with manual grading by ophthalmologists for identifying diabetic retinopathy in retinal fundus photographs.

The algorithm was evaluated at two operating points selected for high specificity and high sensitivity, in two validation sets of 9963 images and 1748 images. The sensitivity and specificity of the algorithm for detecting referable diabetic retinopathy (RDR) defined as moderate or worse diabetic retinopathy or referable macular edema by the majority decision of a panel of at least 7 US board-certified ophthalmologists.

The study found that at the operating point selected for high specificity, the algorithm had 90.3% and 87.0% sensitivity and 98.1% and 98.5% specificity for detecting RDR. At the operating point selected for high sensitivity, the algorithm had 97.5% and 96.1% sensitivity and 93.4% and 93.9% specificity in the 2 validation sets.

The results show that deep learning algorithms had high sensitivity and specificity for detecting diabetic retinopathy and macular edema in retinal fundus photographs. However, further research is necessary to determine the feasibility of applying this algorithm in the clinical setting and whether it could lead to improved care and outcomes.

More articles on My Health Career:

• Guidance for administering chloramphenicol for bacterial conjunctivitis
• The rate of diabetes in patients presenting to emergency deparments
• The patient I almost blinded – a lesson in sticking to my guns – by Amanda Griffiths

Photo credit: National Eye Institute via VisualHunt / CC BY