"FRACTURE” Study
Research type
Research Study
Full title
"FRACTURE” Study: “Fast Reporting using Artificial intelligence for Children’s TraUmatic Radiology Examinations
IRAS ID
274278
Contact name
Susan Shelmerdine
Contact email
Sponsor organisation
Great Ormond Street Hospital for Children NHS Foundation Trust
Duration of Study in the UK
4 years, 10 months, 31 days
Research summary
This study does not require any new data from patients. It uses only imaging data (such as X-rays or CT scans) which have already been performed as part of routine clinical. There is no patient burden.
WHAT IS THE PROBLEM?
Half of all the 12 million children (<16 years) in the UK will break a bone (a fracture) during childhood.
Imaging (typically X-rays or CT) helps doctors see the fracture and make decisions on treatment.Without an expert opinion from a radiologist (a specialist who looks at medical images), mistakes are made in diagnosing 10% of children’s fractures by emergency doctors. Unfortunately, we don’t have enough specialist children’s radiologists in the NHS to check all images immediately, meaning a delay in recognising mistakes.
If fractures are missed, children will be left in pain and won’t get the right treatment, leading to long term problems with discomfort and disability. In some cases, a fracture can be a sign of abuse, and crucial opportunities to arrange for the child’s safety are missed.
WHAT CAN BE DONE?
Artificial intelligence (AI) has the potential to be used to find fractures on imaging as accurately as a radiologist.
AI is a computer programme trained to carry out actions usually done by humans – in this case to find fractures on images.
This could reduce the number of missed fractures, ensure children get the right treatment quickly, reduce inefficiencies for parents and children, and save money for the NHS.A computer programme (using AI) will be developed to ‘learn’ which images are normal and abnormal, using thousands of examples of images from children who previously visited hospitals in the UK. The programme will be tested on images that were not previously shown to the computer programme to see how many mistakes it makes compared to radiologists and assess accuracy.
REC name
N/A
REC reference
N/A