An Evolutionary Federated Learning Approach to Diagnose Alzheimer's Disease Under Uncertainty

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Alzheimer's disease (AD) causes significant cognitive and functional decline, with its etiology still unclear.
• Early AD diagnosis is crucial for interventions to slow disease progression.
• This research addresses AD complexity using multimodal data integration (medical imaging and demographics).

Purpose of the Study:

• To develop a scalable, secure, and privacy-preserving system for Alzheimer's disease diagnosis.
• To integrate multimodal data (MRI, demographics) for improved diagnostic accuracy.
• To evaluate federated learning and belief rule base (BRB) for managing data uncertainty in AD.

Main Methods:

• A deep learning framework using Convolutional Neural Networks (CNNs) processed MRI images.
• Federated learning distributed data for local training across clients.
• A belief rule base (BRB) integrated multimodal data and managed uncertainty during training.

Main Results:

• Federated learning, specifically the FedAvg aggregation method, was evaluated.
• The model achieved a global accuracy of 99.9% in diagnosing Alzheimer's disease.
• The BRB framework effectively handled uncertainty in AD data integration.

Conclusions:

• The developed multimodal approach advances Alzheimer's disease diagnostics.
• Federated learning offers a scalable and privacy-preserving solution for healthcare.
• The BRB framework provides robust data integration and analysis for complex diseases like AD.