Biomedical data repositories require governance for artificial intelligence/machine learning applications at every...

Area of Science:

• Biomedical data science
• Artificial intelligence in healthcare
• Data governance

Background:

• The National Institutes of Health (NIH) launched the Bridge2AI Program to foster AI adoption in biomedical research.
• Four flagship datasets were funded, designed for use with artificial intelligence (AI) and machine learning (ML) technologies.
• Ensuring responsible data collection and governance is crucial for the ethical application of AI in health.

Purpose of the Study:

• To discuss the challenges and lessons learned in collecting and governing AI-ready biomedical datasets.
• To outline the steps and ethical considerations in creating and utilizing these datasets.
• To highlight strategies for responsible data management in AI-driven research.

Main Methods:

• Data selection criteria and justification.
• Incorporation of public concerns and participant consent.
• Establishing protocols for data storage, access, sharing, and downloading.
• Addressing ethical, legal, social, and practical challenges throughout the data lifecycle.

Main Results:

• Identified ethical, legal, social, and practical challenges in AI-ready dataset creation.
• Highlighted diverse project-specific choices in public input, data storage, and access control.
• Demonstrated the importance of addressing future data storage and use concerns.

Conclusions:

• Bridge2AI dataset governance processes, though varied, share common elements.
• These common elements offer valuable strategies for future AI-focused data programs.
• Lessons learned underscore the need for robust data governance frameworks to ensure responsible AI implementation in biomedical research.