An Automated Enhancement System of Diabetic Retinopathy Fundus Image for Eye Care Facilities

Examining retinal fundus images is compulsory for ophthalmologists to spot features of eye diseases. Some problems, including low contrast and blurred retinal fundus image, may seriously affect the diagnostic procedure, leading to misdiagnosis. Low quality of retinal fundus image makes recognition of features for Diabetic Retinopathy (DR) to be harder for the ophthalmologist. Moreover, there is a lack of eye care facilities infrastructure with ophthalmologists to serve multi-morbid DR patients in rural areas, specifically Malaysia. Hence, this research proposed the development of an automated enhancement system based on computer vision and the Artificial Intelligence (AI) field to improve the quality of fundus images captured. The DR Enhancement System (DRES) helps ophthalmologists in the screening aspect by improving the detection of aberrant fundus images. Several methods for improving the fundus image are utilized: Retinex, Contrast Limited Adaptive Histogram Equalization (CLAHE), and Low-light enhancement. Results show that all methods performed better when improving low-contrast and blurred images. This study contributes to the screening process of DR by improving the quality of the retinal fundus image. The developed AI-based system can also help to solve healthcare logistics problems of reaching DR patients in rural areas.