Paediatric Radiology AI: Adoption Faces Key Barriers

AI in paediatric radiology is seeing cautious but growing adoption globally, although its clinical impact remains modest as healthcare organisations continue to face challenges around paediatric data, workflow integration and implementation, according to an international survey of paediatric radiology leaders.

Paediatric radiology encompasses imaging across a wide range of developmental stages, requiring age-specific protocols and interpretation. These factors, together with limited paediatric imaging datasets, create additional challenges for developing, validating and integrating AI tools into clinical practice.

The survey found that while most participating centres had introduced at least one artificial intelligence (AI) tool into clinical practice, many leaders believed further progress would depend less on developing new algorithms and more on improving integration, collaboration and long-term sustainability.

AI Adoption Expands, but Clinical Impact Remains Limited

The cross-sectional survey gathered responses from department leaders and division chiefs at paediatric radiology centres worldwide, with 18 institutions completing the questionnaire, representing a 69% response rate.

Sixteen centres (88.9%) reported implementing at least one AI application. Bone age assessment was the most common use (44.4%), followed by image segmentation and quantification (22.2%), imaging protocol optimisation (16.7%) and natural language processing (16.7%). Despite widespread adoption, respondents rated AI’s average clinical impact at 3.56 out of 5. Only 16.7% described its effect on practice as transformational.

Regional differences were also observed. Natural language processing and reporting applications were concentrated in North America, while centres in the Asia–Pacific region were more likely to describe AI as transformational.

Paediatric Data and Integration Remain Key Obstacles

The biggest barrier identified was the lack of paediatric-specific datasets, cited by 83.3% of respondents. Other commonly reported challenges included integration with existing clinical workflows (66.7%), high costs or unclear return on investment (50%), and cybersecurity concerns (44.4%).

By contrast, the strongest enablers of successful implementation were organisational rather than technical. Vendor maturity and system integration were identified by 72.2% of respondents, while 66.7% highlighted the importance of internal AI champions in driving adoption.

Clinical Integration Over Model Development?

More than half of respondents (55.6%) agreed that paediatric AI research currently places too much emphasis on model development instead of clinical integration, ethics and sustainability.

The findings suggest that addressing paediatric data scarcity through multicentre collaboration, alongside stronger clinical integration and institutional leadership, will be important for ensuring AI is deployed safely, sustainably and effectively across paediatric radiology services.

Reference

Shah H et al. AI integration in pediatric radiology: perspectives from international academic leaders. Eur Radiol. 2026;DOI:10.1007/s00330-026-12728-9.

Featured image: RadVisuals on Adobe Stock