Automated Screening for Diabetic Retinopathy Using Compact Deep Networks

Abstract

Diabetes is a chronic condition affecting millions of people worldwide.One of its major complications is diabetic retinopathy (DR),which is the most common cause of legal blindness in the developedworld. Early screening and treatment of DR prevents visiondeterioration, however the recommendation of yearly screening isoften not being met. Mobile screening centres can increasing DRscreening, however they are time and resource intensive becausea clinician is required to process the images. This process can beimproved through computer aided diagnosis, such as by integratingautomated screening on smartphones. Here we explore the useof a SqueezeNet-based deep network trained on a fundus imagedataset composed of over 88,000 retinal images for the purpose ofcomputer aided screening for diabetic retinopathy. The results ofthis neural network validated the viability of conducting automatedmobile screening of diabetic retinopathy, such as on a smartphoneplatform.