DrStuart Bowyer

Dr Stuart Bowyer's research treats data and AI in healthcare as clinical interventions: things that should be designed, evaluated and monitored like a therapy rather than shipped like a product.

The difficulty is that these interventions do not hold still. The intervention is never the model alone, but the model together with the clinician, the workflow, and the mechanism that updates it. All of these change over time as clinical recording practice, coding and data sources change, and an intervention that performed well in one hospital or one year may not in another. Existing evaluation and reporting standards do not address this, and the methods they would need do not yet exist. Developing them runs through his work, which spans a range of surgical and medical specialties and stages of the pathway from model development to clinical deployment.

His programme is structured around three interlinked areas.

Clinical data science and computational intelligence

He develops computational approaches that turn complex clinical data into usable clinical insight, including interpretable machine learning, large language model reasoning systems, and agentic and neuro-symbolic architectures for structured clinical decision-making. The emphasis is on verification: establishing what a system's outputs can be relied upon to mean before those outputs inform a decision, and making that judgement auditable against clinical standards.

Surgical data science and real-world evidence

Using multimodal data from routine care, his work characterises how patient pathways evolve in practice across surgical and medical specialties, and identifies where data-enabled support would change decisions rather than merely describe them. This includes computable phenotyping in linked NHS datasets, analysis of robotic and operative performance data, and work on how far findings hold when moved between hospitals, where differences in recording convention often matter more than differences in case mix.

Clinical translation and responsible deployment

This area addresses what happens once a tool reaches clinical use. It covers prospective and randomised evaluation of data-enabled interventions, monitoring for drift and performance decay after deployment, bias and equity auditing, and regulatory strategy. He has experience of taking medical devices through regulatory approval and of building analytic platforms inside hospital secure data environments, and works closely with clinical teams and patients on whether new tools are acceptable and safe as well as effective.

He has co-founded three spin-outs translating academic research into hospital and pre-clinical practice, and supervises interdisciplinary research across computer science, engineering and medicine.