WebQuorumChain: A web framework for quorum-based health care model learning.

Area of Science:

• Healthcare Informatics
• Machine Learning
• Data Security

Background:

• Privacy concerns hinder collaborative machine learning in healthcare.
• Decentralized federated learning offers privacy and security but requires complex setups.
• An intuitive, user-friendly, and secure system is needed to improve federated learning utilization.

Purpose of the Study:

• To develop an integrated, web-based system for decentralized federated learning in healthcare.
• To simplify the configuration, management, and monitoring of federated learning networks.
• To promote cross-institutional collaboration while preserving data privacy.

Main Methods:

• Developed WebQuorumChain, an integrated system based on the QuorumChain schema.
• Tested the system on a 2-site network using two public health datasets.
• Measured average vertical and horizontal-ensemble AUCs and execution time over 30 trials.

Main Results:

• Achieved consistently high AUCs (0.94-0.96) across datasets.
• Demonstrated reasonable execution times (5-20 minutes), including system overhead.
• Front-end provides real-time visualization of back-end event logs, synchronized with algorithm progress.

Conclusions:

• WebQuorumChain provides visual tools for network configuration, training management, and process inspection.
• The system facilitates secure, low-level process monitoring without compromising data privacy.
• It maintains predictive accuracy and efficiency, promoting data-driven discoveries in healthcare through collaboration.