System for Classification and Diagnosis Assistance of Diabetic Retinopathy

System for Classification and Diagnosis Assistance of Diabetic Retinopathy

China Medical University Hospital

Retinopathy is an important complication of diabetes. In Taiwan, first-line care division for diabetic patients are family medicine, metabolism clinics. Fundus photography must be arranged for diabetic patients to screen symptoms of retinopathy. The image interpretation is extremely dependent on the experience of ophthalmologists. In response to physicians who are not ophthalmologists, this system provides relevant auxiliary decision-making information to assist physicians in subsequent processing.

This system is assisted by a team of professional ophthalmologists to label a large number of normal and diabetic retinopathy images as a training data set to develop a deep neural network structure. Machines learn the characteristic patterns of normal images and abnormal images caused by lesions to assist in diagnosis and classification. It also communicates closely with clinicians to verify, increase training data and improve accuracy rate through instant feedback from physicians, eventually to achieve a smart medical ecological cycle.

Source：2020 Application of Smart Medical Technology brochure