Federated Deep Learning Approaches for Detecting Ocular Diseases in Medical Imaging: A Systematic Review

Area of Science:

• Ophthalmology
• Computer Science
• Medical Imaging

Background:

• Artificial intelligence (AI), particularly Deep Learning (DL) and Federated Learning (FL), is revolutionizing healthcare diagnostics.
• FL offers a privacy-preserving solution for analyzing sensitive medical data, crucial for detecting ocular diseases from medical images.
• This review systematically examines AI advancements in ocular disease detection, emphasizing FL's role in enhancing accuracy and confidentiality.

Purpose of the Study:

• To conduct a systematic review of AI-driven ocular disease detection methods.
• To specifically analyze the evolution, methodologies, and effectiveness of Federated Learning (FL) in this domain.
• To evaluate FL's ability to ensure data privacy while maintaining diagnostic performance.

Main Methods:

• Systematic review adhering to the PRISMA framework.
• Literature search conducted across major academic databases (Web of Science, Scopus, IEEE Xplore, PubMed) from 2017-2024.
• Selection of studies focused on DL and FL models for ocular disease detection, followed by comparative analysis of methodologies, datasets, and performance.

Main Results:

• FL models demonstrate diagnostic performance on par with traditional centralized AI approaches.
• High accuracy and efficiency reported for FL models like FedAvg and FedProx in ocular disease detection.
• Key challenges identified include data heterogeneity, communication efficiency, and model convergence.

Conclusions:

• Federated Learning (FL) is a promising strategy for ocular disease detection, effectively balancing diagnostic accuracy with robust data privacy.
• Ongoing research is essential to optimize FL frameworks for scalability and communication efficiency.
• Integration of advanced privacy-preserving techniques is a critical future direction for FL in medical diagnostics.