Why Unilabs is supporting its radiology peers to implement and monitor AI tools

Artificial intelligence (AI) is rapidly reshaping radiology, promising greater diagnostic accuracy, faster workflows, and improved patient care. Introducing AI into clinical practice is, however, not as simple as deploying a tool and letting it run. It requires oversight, continuous monitoring, and close collaboration between radiologists, data scientists, and operational teams.

In their latest peer-reviewed research paper, Real-World Monitoring of Artificial Intelligence in Radiology: Challenges and Best Practices­—cited by the Royal College of Radiologists in its AI monitoring guidance—Unilabs Radiologist and AI Clinical Lead, Dr. Geraldine Dean, and Radiologist and Head of Unilabs Radiology, Dr. Gareth Davies, explain the critical importance of monitoring AI. They also share practical insights to ensure the safe and ethical implementation of AI tools for the benefit of healthcare professionals and patients around the world.

Implementing AI is only half the story

An increasing number of healthcare organisations, including Unilabs, are publishing guidance around implementing AI tools to improve diagnostic accuracy, fast-track triaging, and boost workflow efficiency. Unilabs is also taking the next step and ensuring the safe use of AI post-implementation through the delivery of continuous monitoring programmes.

Dr Dean believes there is a need for more in-depth discussions and real world studies around post-implementation monitoring, and says it was this conviction that drove a group of Unilabs radiologists, data scientists, and the AI Centre of Excellence (AICoE) team to publish learnings on the importance of structured monitoring programmes. “Unilabs has been systematically integrating AI into radiology and pathology workflows across multiple countries since 2016. While much of this work has been internal, we recognise that the insights, frameworks, lessons learnt, and challenges we’ve navigated could help the broader community.”

She says monitoring the use and output of AI  tools and the human-AI interaction is a fundamental part of ensuring that AI implementation is ethical,  safe, and creating true value.  “AI tools are only beneficial if they are working as expected, for their intended purpose, and are being used responsibly. With regular, structured monitoring, organisations can ensure this is the case.”

Dr Dean says structured monitoring programmes can identify and alert to any issues and, crucially, put in place processes to mitigate any risks. This ensures that patients receive an accurate diagnosis, whilst radiologists feel supported, and AI systems continue to deliver real value. 

A multidisciplinary approach

Embedding and monitoring AI tools in radiology and pathology requires expertise, insight, and feedback from a wide range of experts. At Unilabs, the AICoE team works closely with clinical, operations, and legal teams, as well as AI engineers, data scientists, and IT and business experts. These teams also work together with the radiology research team—a community of research-driven radiologists and data scientists—to ensure that every aspect of AI implementation is grounded in evidence and can be used to refine workflows, guide best practice, and support safe, effective AI deployment in clinical care.

Dr Davies says, so far, Unilabs has implemented 15 AI tools across six countries. “We’re uniquely positioned to lead the conversation on AI in radiology and pathology. We ensure that our radiologists and pathologists remain at the forefront of AI-driven innovation. They help us to identify genuine clinical needs, provide continuous feedback, and reassess tools to ensure safety and relevance. Our AI solutions are therefore shaped by our clinicians, ensuring that technology serves clinical needs, not the other way around.”

He says through multidisciplinary collaboration, a vast and diverse data footprint, and the continuous learning fostered by the Unilabs Academy, Unilabs empowers radiologists and pathologists to be innovators and co-creators of AI tools, not just users.