Responsible Artificial Intelligence Governance in Oncology
Artificial Intelligence (AI) is transforming healthcare, with oncology being one of the most impactful areas of application. This study explores the governance frameworks necessary for the responsible implementation of AI technologies in oncology, focusing on their potential effects on patient care and operational efficiency.
Introduction
AI is increasingly being integrated into various stages of a cancer patient’s journey, including clinical trial matching, care team decision support, and the broader spectrum of cancer research. As of August 2024, the FDA’s registry lists 949 AI-enabled medical devices, with 75 specifically targeting oncology. This growth underscores the critical need for governance frameworks tailored to the unique challenges and opportunities presented by AI in cancer care.
The Need for Responsible AI (RAI) Frameworks
Despite the proliferation of AI technologies, oncology-specific frameworks for Responsible AI governance are still in their infancy. While general RAI guidelines are emerging, none adequately address the specific needs of oncology. This is crucial given the unique challenges of ensuring equitable cancer care amidst existing disparities in access and treatment outcomes.
Framework Development and Implementation
This study outlines a two-phase approach to developing and implementing an RAI governance framework in oncology:
Phase 1: Design and Development
This phase involved the establishment of an AI Task Force (AITF) to identify the main challenges in AI governance, such as:
- High-quality data for model development.
- Computing power requirements.
- AI talent capacity.
- Policies and procedures.
The AITF assessed the current state of AI use in oncology, identifying 87 active AI projects across research, clinical, and operational domains. This resulted in the formation of strategic goals aimed at enhancing AI governance.
Phase 2: Implementation
The second phase focused on implementing the governance framework, including:
- The establishment of an AI Governance Committee (AIGC) to oversee AI model deployment.
- Development of a Model Information Sheet (MIS) to register all AI models.
- Implementation of a risk assessment framework for evaluating AI models before deployment.
The AIGC’s structure is embedded within existing digital governance systems, ensuring integration without stifling innovation. This dual approach aims to balance the promotion of AI technologies and their responsible use.
Real-World Applications and Case Studies
To illustrate the governance framework in action, the study presents two case studies:
- The first involves an acquired AI model for mammogram triage developed by a third-party vendor, demonstrating how the AIGC facilitated its expedited review and deployment.
- The second focuses on an internally developed model for tumor segmentation in brain metastases, showcasing the AIGC’s role in ensuring thorough evaluation and ongoing monitoring.
These cases highlight the importance of maintaining rigorous standards while promoting innovation in AI applications within oncology.
Conclusion
This study emphasizes that effective governance of AI in oncology is not only feasible but essential. As AI projects increase, the need for tailored frameworks that address the unique demands of oncology becomes more critical. By adopting a comprehensive RAI governance model, healthcare organizations can ensure that AI technologies enhance patient care while adhering to ethical standards.
Overall, the implementation of RAI frameworks in oncology presents a significant opportunity to improve health outcomes and operational efficiencies while addressing the ethical considerations inherent in the deployment of AI technologies.
