UK researchers at Imperial College have teamed up with the Children’s Hospital of Orange County (CHOC) in California to develop AI support systems for paediatric intensive care.

The researchers, aided by clinicians, are training an ‘AI Clinician’ model to help paediatric doctors make complex decisions and provide faster and more accurate treatment recommendations for critically ill children.

The project, led by Professor Aldo Faisal, Director for AI for Healthcare Centres and Imperial Global USA, has brought together researchers and clinicians from Imperial’s Department of Bioengineering, the Department of Computing and the Department of Surgery and Cancer, with a team of data scientists and clinicians at CHOC. The initial research on this project was funded by Rosetrees as a multidisciplinary award to develop an AI tool for personalised care for children with sepsis.

Unlike earlier models that relied on simple rules, Imperial’s AI Clinician model uses advanced algorithms that consider the full complexity of a patient’s health and subsequently provides treatment recommendations.

The UK/US partnership has combined over 56,000 anonymised records from the NHS and CHOC to create the largest international paediatric dataset of its kind, with the aim of training the AI Clinician to help with a broad diversity of conditions and medical backgrounds.

Such a model may be crucial in analysing the huge amount of patient data that is generated in a typical paediatric intensive care unit (up to 100 dataset categories every hour), helping clinicians make treatment decisions, quickly and effectively in high pressure and life-threatening situations where each patient is unique and slight variations in treatment can impact recovery.

“By combining granular data from tens of thousands of children admitted to paediatric critical care units in the UK and USA, this effort has the potential to revolutionise care for sick children in the future,” said Dr Padmanabhan Ramnarayan, Deputy Director of Imperial’s Centre for Paediatrics and Child Health.