Building AI Capability in Medical Imaging: A Co-Designed Continuing Professional Development Framework and Evaluation

Background: Artificial intelligence (AI) is increasingly embedded in medical imaging workflows, yet many imaging professionals report limited preparation to evaluate, implement, and govern AI tools safely. This educational gap risks inappropriate reliance on AI systems and undermines effective clinical oversight and patient safety.

Methods: We undertook a Three-phase mixed-methods study to co-design and evaluate a tiered AI education framework for healthcare professionals, with an emphasis on medical imaging. A hybrid co-design workshop involving 52 healthcare stakeholders identified AI knowledge gaps, role-specific needs, and training preferences, informing a Three-pathway framework spanning foundational AI literacy, imaging-focused proficiency, and policy and governance. Two continuing professional development courses - AI Literacy in Healthcare and AI in Medical Imaging – were subsequently designed and delivered to over 300 healthcare professionals. Pre- and post-course surveys (132/59 responses for AI literacy; 26/30 for AI in medical imaging) captured self-reported changes in knowledge, confidence, and understanding of ethical and regulatory issues; quantitative data were analysed descriptively and qualitative free-text responses thematically.

Results: Workshop participants reported widespread gaps in foundational AI literacy, critical appraisal skills, and awareness of governance and regulatory requirements, and strongly endorsed the need for structured, role-specific training. For the AI literacy course, participants reported substantial short-term gains in core AI concepts, clinical use cases, and responsible AI principles, alongside increased confidence in discussing AI with colleagues and patients. Among imaging professionals, the AI in medical imaging course was associated with marked perceived improvements in understanding AI model training and validation, performance metrics, AI explainability, bias, and workflow integration with PACS/RIS, and in readiness to engage with AI-enabled imaging tools under appropriate human oversight.

Conclusion: A co-designed, tiered AI education framework can address heterogeneous AI literacy and capability needs across the healthcare and medical imaging workforce, strengthening confidence and perceived preparedness for safe AI adoption. The proposed pathways offer a clinically grounded, governance-aware, and scalable structure that healthcare organisations and educators can adapt to support progressive, role-aligned AI capability building in medical imaging and related domains.