A team of Japanese researchers have developed a deep learning algorithm to automatically diagnose glaucoma with a fundus camera.
The study used 1,364 color fundus photographs of eyes that exhibited signs of glaucoma and 1,768 color fundus images of normal eyes as the dataset. Testing consisted of 34 eyes with non-highly myopic glaucoma, 30 eyes with highly myopic glaucoma, 28 non-highly myopic eyes of normal subjects and 22 highly myopic eyes of normal subjects. Diagnostic accuracy was validated using several statistical measures.
The paper is being presented today at the annual ARVO conference currently underway in Honolulu, Hawaii.
“Of interest, color fundus photographs employing a deep learning algorithm has been developed to diagnose glaucoma damage and shows a very high diagnostic ability,” says Joseph P. Shovlin, OD, of Scranton, PA. “This may aid the clinician by simply employing the deep learning algorithm which provides good sensitivity and specificity using a fundus photograph.”
Researchers also found their algorithm had a high diagnostic ability both in non-highly myopic and highly-myopic eyes.
|Asaoka R, Shibata N, Murata H, et al. Construction of a deep learning algorithm to automatically diagnose glaucoma using a fundus photograph. ARVO 2018. Abstract 3024.|