Privacy-aware vaccine recommendation using federated learning and blockchain

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Blockchain Technology

Background:

• Existing vaccine recommendation systems lack personalization based on individual symptoms and medical history across diverse geographical locations.
• Hyperledger Fabric blockchain has not been fully utilized to integrate user data for enhanced vaccine recommendations.

Purpose of the Study:

• To develop an advanced vaccine recommendation system integrating user symptoms and medical history.
• To implement a federated learning approach for training patient data across various locations securely.
• To enhance vaccine accessibility and personalization through a novel AI and blockchain-based system.

Main Methods:

• Utilized Krichevsky Dirichlet Trofimov-based latent Dirichlet allocation (KDT-LDA) for symptoms and medical history modeling.
• Employed federated learning with Expcos bidirectional distillation long short-term memory (FL-EBiDLSTM) for personalized vaccine prediction.
• Integrated Hyperledger Fabric blockchain for secure, real-time user data management and system registration.

Main Results:

• The proposed KDT-LDA and FL-EBiDLSTM model achieved a high accuracy of 99% in vaccine prediction.
• The system effectively processed and modeled symptoms and medical history using advanced NLP techniques.
• Federated learning successfully addressed the challenge of training patient data across multiple geographical locations.

Conclusions:

• The developed vaccine recommendation system demonstrates superior performance and accuracy compared to existing methods.
• The integration of AI and blockchain technology offers a secure and efficient platform for personalized healthcare solutions.
• This approach significantly improves the potential for personalized vaccine recommendations, enhancing public health outcomes.