Research has recently suggested that ethnicity and race may be significant factors in the diagnostic accuracy of ocular imaging techniques.^1,2 Certain OCT parameters such as circumpapillary capillary density differ between races, and OCT databases that are ethnicity specific could improve the tool’s specificity for glaucoma detection.^1,2

Now, a new study suggests a deep learning algorithm based on OCT images can correctly identify AMD features regardless of ethnicity.³ Researchers in South Korea found their algorithm appropriately focused on the differences within the macular area to extract features associated with wet AMD—findings they validated in their own population and in an ethnically diverse population in the United States.

The model’s development data set included 12,247 OCT scans from South Korea. After internal testing, the algorithm was sent to the University of Washington in Seattle for external validation with 91,509 OCT scans.

On external validation, the algorithms accuracy remained high at the individual level. Further analysis showed that, in normal OCT scans, the fovea was the main area of interest. In wet AMD OCT scans, the model noted the appropriate pathological features, specifically subretinal fluid and pigment epithelial detachment. A survey of 10 retina specialists confirmed the model was appropriately identifying lesions indicative of wet AMD on the OCT scans.

The researchers were pleased to find the deep learning algorithm performed well for wet AMD identification in a Korean population, and it translated well to the ethnically distinct American population. They note that the next challenge is helping clinicians and patients understand the algorithm’s reasons for disease classification and diagnosis.