Diabetic retinopathy screening using machine learning: a systematic review

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of blindness globally.
• Manual screening of retinal images is time-consuming and faces expert shortages.
• Machine learning (ML) and deep learning (DL) present viable alternatives for DR screening.

Purpose of the Study:

• To analyze the research landscape of ML integration in diabetic retinopathy screening.
• To identify and characterize available retinal fundus image datasets.
• To discuss preprocessing techniques, ML progress, challenges, and future directions in DR screening.

Main Methods:

• Literature review and analysis of ML techniques applied to DR screening.
• Characterization of publicly available retinal fundus image datasets.
• Discussion of common image preprocessing methods for DR detection.

Main Results:

• Identified and characterized available retinal image datasets for DR screening.
• Analyzed the progress and application of various ML techniques in DR detection.
• Highlighted common preprocessing steps essential for effective DR screening.

Conclusions:

• ML integration shows significant potential to improve the efficiency and accessibility of diabetic retinopathy screening.
• Standardized datasets, model complexity, and computational resources remain key challenges.
• Further research is needed to overcome existing hurdles and advance ML-based DR screening solutions.