Heartbeat: Radiology leading way in artificial intelligence
July 20, 2022
Three years ago, Group Director of Imaging Dr Sarah Yusuf realised that the use of Artificial Intelligence (AI) in radiology had the potential to transform how we care for our patients and report results following imaging. AI is becoming more common in healthcare and our Trust is doing some exciting work in this field of research.
Sarah said: “Our radiology team has been working with IBM, our imaging system provider who have developed their Watson artificial intelligence system, they wanted a European partner to validate their CR3 AI program, which looked at reported chest x-rays and CTs, to ensure that six pre-identified findings were correctly reported.
“My colleague Dr John Morlese and I, with the support of Physicist, Bahadar Bhatia, with the support of IT thanks to Martin Sadler, undertook this validation process. We sent 13,000 images to the AI algorithm and reviewed all cases where the algorithms felt the radiologist had missed the finding.
“Reassuringly, our radiology reporting error rate is remarkably low, 0.7% in the study, which is well below the RCR threshold for acceptable levels of errors.
“Some of the issues with the algorithm related to its inability to recognise the different ways in which radiologist described lesions, and so our feedback provided data to allow tweaking of the neurolinguistic component of the software and a new version 3.3 is now available commercially.
We have forged a close relationship with IBM and are looking forward to working on future algorithms that characterise lesions in prostate & liver MRIs’
Martin Sadler, Executive Director of IT and Digital added: “The results of our tests have been extremely positive, firstly we have shown that we have skilled and effective radiologists, which we already knew really, but also we have trained the artificial intelligence system to be as accurate on a variety of studies.
“The three years of work have also been complemented and supported by the steady improvement in IT services allowing us to be able to innovate – we have moved all our imaging services into the cloud, ensuring that we now have a reliable and resilient infrastructure that can grow with us.”
Sarah added: “We are confident that there will be tangible benefits to patient care as a result of our work and investment in AI, particularly in helping identify high and low risk studies, to prioritise them and to manage workload. Our initial findings are currently being finalised and we intend to publish them.
“AI learns from the data with which it is presented, these systems are incredibly skilled at pattern matching and detecting things that could otherwise be overlooked. We are hoping that both staff and patients will soon be able to reap the benefits of AI, supporting us to improve accuracy and reporting times. The accuracy, understandings and speed of a focussed AI system that has a good dataset is unmatched.
“We continue to explore other ways in which AI can help us in the Trust and are grateful for colleagues within imaging and informatics who are working to ensure we get this right for patients. We are particularly delighted to have Bahadar on board, he has been deeply involved in the progress and has pioneered a number of other studies looking into a variety of image related Artificial Intelligence.”
Earlier this year, the team hosted a conference for IBM at Sandwell Hospital aimed at raising awareness and attracting external funding.
Sarah and Martin have also attended various healthcare forums to showcase the initial findings of the study, explain the IT journey that is required to support AI activity and discuss our organisation’s AI ambitions.