LLM-Assisted Scoping Review of Artificial Intelligence in Brazilian Public Health: Lessons from Transfer and...

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Global Health Technology

Background:

• Artificial intelligence (AI) is a critical technology for global health, especially concerning climate change and digital divides.
• AI applications in healthcare are increasingly important for addressing health system challenges.
• Brazil's healthcare landscape presents unique opportunities and challenges for AI integration.

Purpose of the Study:

• To conduct a scoping review of AI applications in Brazilian healthcare.
• To specifically analyze the use of Transfer Learning (TL) and Federated Learning (FL) in this context.
• To understand how TL and FL address data scarcity, privacy, and technological dependence in AI healthcare solutions.

Main Methods:

• A comprehensive scoping review was performed, searching databases like PubMed, SciELO, and CNPq.
• AI-assisted title screening with manual validation was employed.
• Thematic analysis focused on methodological and infrastructural aspects of AI applications.

Main Results:

• Out of 349 retrieved studies, only six explicitly utilized Transfer Learning (TL) or Federated Learning (FL).
• TL and FL were often implemented within multi-country research consortia, highlighting their collaborative potential and scalability.
• Despite their benefits for privacy-preserving, resource-efficient AI deployment, these methods remain underutilized in mainstream Brazilian healthcare.

Conclusions:

• The integration of resource-aware AI, including TL and FL, in Brazil's health sector is nascent and uneven.
• These AI approaches offer significant potential for equitable innovation and digital autonomy in the Global South's health systems.
• Further adoption of TL and FL is crucial for advancing AI capabilities while respecting data sovereignty and addressing digital inequalities.