23 January 2019
AI system can speed up prioritising patient chest X-rays
Research from King's, supported by the NIHR Guy’s and St Thomas’ Biomedical Research Centre (BRC), has shown that AI can dramatically reduce the time it takes to ensure abnormal chest X-rays receive expert radiologist opinion.
Chest X-rays are routinely performed to diagnose and monitor a wide range of conditions affecting the lungs, heart, bones, and soft tissues. Ensuring that abnormal X-rays showing signs of urgent illness are reviewed by a radiologist as soon as possible, will ensure patients get the most appropriate care promptly and improve treatment outcomes.
The researchers, from the School of Biomedical Engineering and Imaging Sciences at King’s and the University of Warwick, trained an algorithm to recognise abnormalities in chest X-rays, using a dataset of 500,000 anonymised adult chest X-rays from patients at Guy’s and St Thomas’, which had been assessed by expert radiologists.
By applying this algorithm to such a large number of chest X-rays, the AI system was able to interpret the visual patterns on the X-rays, predict their urgency and suggest a priority level for the X-rays to be reported by a radiologist.
The team then tested the AI system in a simulation and found it could cut the average reporting time from 11 days to less than three days. Treatment outcomes are therefore likely to be improved due to reduction in reporting delays. They said that AI could help ensure that patients with X-ray abnormalities would be seen and treated sooner in the future.
Study author, Professor Vicky Goh, Professor of Cancer Imaging at King’s said: “Chest X-rays make up over half of the world’s imaging currently. Using AI to detect normal from abnormal X-rays accurately and speeding up the treatment of patients with significant illness, is a real game-changer for patients worldwide. We are looking forward to working with Professor Giovanni Montana’s team at Warwick, to improve the accuracy of this system and to test it in multiple hospitals, funded by the Wellcome Trust.”
The full paper is available online.