Advances in disease detection through retinal imaging: A systematic review

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Ocular and non-ocular diseases cause significant vision impairment globally.
• Early detection and management are crucial for preventing blindness.
• Manual disease detection from retinal images is labor-intensive and experience-dependent.

Purpose of the Study:

• To systematically review machine learning (ML) techniques for disease detection from retinal images.
• To analyze the efficiency of ML models in single and multi-modal imaging.
• To identify challenges and propose future research directions in ML-based ocular diagnostics.

Main Methods:

• Systematic literature review of ML applications in retinal image analysis.
• Analysis of Deep Learning and classical ML models for disease detection.
• Evaluation of model performance based on accuracy, sensitivity, and specificity.

Main Results:

• ML techniques show promise for automated disease detection and grading.
• Both single and multi-modal imaging approaches are utilized.
• Various ML models demonstrate significant achievements in diagnostic performance.

Conclusions:

• ML enhances diagnostic accuracy and patient outcomes in ocular disease management.
• Addressing identified challenges is key to realizing ML's full potential.
• Future research should focus on overcoming limitations for reliable disease diagnosis.