Medical Imaging Applications of Federated Learning.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computer Science

Background:

• Federated Learning (FL) offers inherent security, privacy, and scalability advantages.
• FL's ability to overcome data biases makes it suitable for sensitive healthcare datasets.
• Existing reviews cover FL broadly, but a focused review on medical imaging applications is needed.

Purpose of the Study:

• To systematically review Federated Learning applications in medical imaging.
• To categorize FL applications by disease, imaging modality, and anatomical part.
• To analyze FL model performance against traditional Machine Learning models.

Main Methods:

• Systematic literature review of ArXiv, IEEE Xplorer, and PubMed.
• Detailed description of FL architectures and models used in medical imaging.
• Comparative analysis of FL and traditional Machine Learning performance.

Main Results:

• FL is increasingly applied to diverse medical imaging tasks, showing promising results.
• Privacy-preserving techniques like homomorphic encryption and differential privacy are key FL components.
• Performance of FL models is comparable to traditional ML, with ongoing improvements.

Conclusions:

• Federated Learning shows significant potential for secure and private medical imaging analysis.
• Security and data-related challenges remain primary concerns for FL deployment in healthcare.
• Further research and development are crucial for advancing FL in medical imaging.