AI in Biomedicine: Navigating Challenges for Global Impact

By Soham NandiReviewed by Susha Cheriyedath, M.Sc.Jan 8 2024

In an article published in the journal PNAS Nexus, researchers discussed the transformative potential of artificial intelligence (AI) and machine learning (ML) in biomedical research and healthcare. It emphasized the importance of responsible, ethical, and equitable AI and ML integration, addressing challenges, and promoting inclusivity.

Background

The paper explored the profound impact of AI and ML on biomedical research and healthcare, envisioning improvements in operational efficiency, cost reduction, enhanced diagnostics, and personalized treatment. Acknowledging the challenges, it emphasized the need for responsible implementation, aligning with President Biden's executive order on AI standards. However, existing issues include workforce disparities, geographic concentration of innovation, biases in datasets, and potential post-approval inequities.

The government and stakeholders worked on oversight frameworks, but private industry set the pace due to high entry barriers. The National Academies aimed to bridge gaps by convening stakeholders, providing evidence-based recommendations, and fostering discussions. Previous workshops focused on AI's potential in scientific discovery and healthcare but revealed infrastructure gaps, governance needs, and opportunities for international collaboration. This paper sought to address these issues, advocating for a mission-driven approach prioritizing social good over profit-driven motives in the future of AI and ML integration for health and research.

Infrastructure Requirement

The researchers underscored the imperative need for robust infrastructures in biomedical science, particularly in the context of AI's potential in personalized and precision medicine. They emphasized the necessity of democratizing access to research and outcomes, focusing on four key types of infrastructure.

• Infrastructure for Data: Requires a commitment to collecting, curating, and managing diverse data reflecting all communities. Innovative tools ensuring data quality and privacy-preserving techniques were emphasized to address the growing risks associated with increased data consumption by AI models.
• Infrastructure for computation: Stresses the necessity for a public-interest research infrastructure for AI in healthcare. Accessibility was deemed a top priority, with tools, interfaces, workflows, and tailored training materials being essential to empower researchers across the scientific spectrum.
• Infrastructure for health: Recognizes the current hoarded nature of health data within various systems. The goal was to integrate datasets comprehensively, addressing regulatory challenges and requiring significant funding to achieve interoperability across different data types and settings.
• Infrastructure to scale impact: Emphasizes transparency in AI models to ascertain quality and the need for a professional workforce. The researchers advocated for a global collaboration in AI investments for medicine, fostering accessibility, equity, and rights preservation while mitigating risks.

In essence, the proposed infrastructural enhancements aimed to democratize access to AI-driven biomedical advancements, ensuring inclusivity, transparency, and ethical considerations in improving healthcare outcomes.

Creating Dynamic and Equitable Governance

The authors emphasized the pressing need for dynamic and equitable governance in the rapidly advancing field of AI and ML, particularly within the context of healthcare. While the emergence of powerful AI models and smart technologies brought global accessibility, a comprehensive governance structure was lacking at both national and international levels. The US Office of Science and Technology Policy's Blueprint and the recent executive order offered initial steps, but a more comprehensive approach was needed.

The Collingridge dilemma was highlighted, emphasizing the challenge of predicting the impact of technology early in development and the potential lateness of policymaking during late-stage development or post-approval. The proposed solution involved iterative and dynamic governance, capable of capturing the evolving landscape of AI across sectors and stages of development. Equitable innovation in healthcare and medicine requires a multifaceted governance approach, encompassing both "hard" governance with legally binding laws, and "soft" governance with voluntary guidelines and societal standards.

Global collaboration within the scientific community was deemed essential. Achieving the right balance of incentives and guardrails, prioritizing equity-driven decision-making, and incorporating technology transfer and intellectual property considerations were identified as key components in shaping effective governance frameworks for AI and ML in healthcare.

Building international collaboration and capabilities

The researchers underscored the importance of international collaboration in advancing AI-related scientific efforts, especially within the context of complex global health challenges. Recognizing the interconnected nature of contemporary health issues, the authors advocated for collaborative, all-encompassing approaches to address threats such as pandemics, climate change, and systemic inequities.

Current US federal initiatives leading the AI Safety Summit in the UK and participation in global organizations like the Organization for Economic Cooperation and Development AI Policy Observatory and the Global Partnership on AI were acknowledged. However, the authors called for a more comprehensive vision and stronger health perspective in international AI collaboration discussions, emphasizing the need for diverse and inclusive global efforts.

Conclusion

In conclusion, realizing the transformative potential of AI in healthcare necessitated comprehensive efforts. The focus must be extended beyond addressing current challenges to include building robust infrastructure, dynamic governance, and fostering international collaboration. Equity, oversight, and regulation are vital, but attention to infrastructure development, mission-driven governance, and global cooperation is equally crucial. A holistic approach is essential to harness AI's benefits, ensuring inclusivity and efficiency in addressing complex societal health challenges. Immediate action is imperative to establish a strong foundation for AI-driven growth and discovery.