Abstract

Info

Title:: Classification and detection method for diabetic retinopathy based on the combination of improved Retinex image enhancement and deep learning

Author(s):: WANG Wenjing1; ZHANG Lichuan1; WANG Xin1; LIU Yuhong1; 2; 1. Department of Biomedical Engineering, Chengdu Medical College, Chengdu 610500, China 2. School of Information and Software Engineering, University of Electronic Science and Technology of China, Chengdu 610054, China

Keywords:: Keywords: deep learning image classification image enhancement diabetic retinopathy

Abstract:: Abstract: Objective To present a novel method based on the image enhancement algorithm and deep learning for automatically classifying diabetic retinopathy images, and realizing the graded classification of fundus lesions. Methods An improved Retinex image enhancement algorithm was employed to preprocess the original images for significantly improving image quality and visual effect, and enhancing image clarity and contrast. Then, deep learning method was used to automatically detect and classify the degree of lesions in different periods. Results The proposed method was advantageous in improving classification accuracy, sensitivity, and specificity which were 5.4%, 7.4%, and 16.6% higher than those of traditional Retinex method. Conclusion The proposed method can effectively realize the automatic detection and classification of diabetic retinopathy, which is helpful to enhance diagnostic accuracy and efficiency.