Cloud computing for equitable, data-driven dementia medicine

Area of Science:

• Neurology
• Artificial Intelligence
• Health Informatics

Background:

• Dementia presents a growing global health challenge, necessitating advanced predictive models for new drug development and clinical application.
• Machine learning (ML) shows potential in dementia research but faces significant barriers in routine healthcare, primarily due to data unavailability and resulting data drift.
• Existing ML applications in healthcare are underutilized, especially for image-based decision support, hindering equitable real-world translation.

Purpose of the Study:

• To propose and pilot a scalable, cloud-based infrastructure as code solution for dementia research.
• To address data unavailability and data drift challenges in developing AI models for dementia care.
• To enable the creation of robust and adaptable artificial intelligence (AI) models for dementia medicine while preserving patient privacy.

Main Methods:

• Implementation of a cloud-based infrastructure as code (IaC) solution.
• Integration of privacy-preserving federated learning (FL) techniques.
• Piloting the architecture to demonstrate its scalability and effectiveness in a healthcare context.

Main Results:

• Demonstrated a scalable, cloud-based infrastructure for AI model development in dementia research.
• Successfully implemented privacy-preserving federated learning, keeping patient data localized and secure.
• Enabled the development of adaptable AI models, overcoming data drift barriers.

Conclusions:

• Cloud-based federated learning offers a viable solution for developing AI in dementia medicine.
• This approach enhances data security and patient privacy, crucial for healthcare applications.
• The proposed infrastructure and codebase can accelerate research and real-world translation of AI in dementia care.