Systematic Review of Privacy-Preserving Distributed Machine Learning From Federated Databases in Health Care

Area of Science:

• Health Informatics
• Machine Learning
• Data Privacy

Background:

• Big data presents solutions for healthcare challenges like rising costs and noncommunicable diseases.
• Data centralization for big data in healthcare raises significant privacy and regulatory concerns.
• Distributed learning emerges as a privacy-preserving alternative to traditional data centralization.

Purpose of the Study:

• To review major implementations of distributed learning in healthcare.
• To explore distributed learning as a solution for big data privacy concerns in medicine.
• To examine the potential of distributed learning for international medical research consortiums.

Main Methods:

• Systematic review of distributed learning implementations in healthcare.
• Analysis of search results concerning federated databases and distributed algorithms.
• Examination of legal contexts for patient data research and machine/deep learning concepts.

Main Results:

• Distributed learning enables analysis of separate, isolated datasets without data centralization.
• Algorithms share research questions and answers between databases, not raw patient data.
• This approach facilitates big data utilization in healthcare, especially for international collaborations.

Conclusions:

• Distributed learning is a promising solution for leveraging big data in healthcare while upholding patient privacy.
• It addresses regulatory concerns associated with centralized health data.
• Further review of its implementations is crucial for advancing medical applications.