Human-AI interaction and collaboration in radiology: from conceptual frameworks to responsible implementation

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) is increasingly used in radiology.
• Current studies often evaluate AI algorithms in isolation, not within clinical workflows.
• Understanding human-AI interaction is crucial for effective integration.

Purpose of the Study:

• To review current knowledge on human-AI interaction in radiology.
• To highlight practical risks and opportunities for clinical teams.
• To guide responsible AI implementation in radiology.

Main Methods:

• Narrative review of existing literature on human-AI collaboration in radiology.
• Analysis of AI integration strategies across the imaging pathway.
• Discussion of cognitive, professional, and implementation aspects.

Main Results:

• Conceptual models like diagnostic complementarity explain synergistic performance.
• AI integration strategies span acquisition, triage, interpretation, reporting, and teaching.
• Potential effects include automation bias, deskilling, and burnout, with specific trainee vulnerabilities.

Conclusions:

• Responsible AI implementation requires addressing liability, oversight, performance monitoring, and AI literacy.
• Co-design with radiology teams and usable explanations are key.
• Emerging systems like vision-language models and adaptive learning loops show future potential.