Review on diabetic retinopathy with deep learning methods.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss, necessitating early detection.
• Traditional DR diagnosis relies on ophthalmologists and is often complex due to early asymptomatic stages.
• Automated DR detection using image processing, machine learning (ML), and deep learning (DL) shows promise.

Purpose of the Study:

• To review existing literature on ML and DL techniques for diabetic retinopathy recognition.
• To address challenges associated with datasets used in DR detection studies.
• To provide a comparative analysis of current methods and identify future research directions.

Main Methods:

• Systematic literature review of studies employing ML and DL for DR detection.
• Comparative analysis of various DR datasets, performance metrics, and ML/DL techniques.
• Identification of technical and clinical challenges in DR detection.

Main Results:

• A comparative analysis of databases, performance metrics, and ML/DL techniques used in recent DR detection studies.
• Evaluation of the strengths and limitations of current automated DR detection methods.
• Identification of key challenges in DR dataset utilization and model performance.

Conclusions:

• The review synthesizes current ML and DL approaches for diabetic retinopathy detection.
• It highlights critical technical and clinical challenges often overlooked in existing reviews.
• Provides insights into future research scopes for advancing retinal imaging analysis in DR.