Oxford study finds AI can be safely used to help diagnose prostate cancer
The technology could make diagnoses more accurate and reduce pressure on labs
Artificial intelligence could help doctors diagnose prostate cancer more accurately while speeding up results for patients, according to a landmark Oxford-led study.
Researchers tested AI software alongside specialist histopathologists examining more than 1,000 prostate biopsy samples across three NHS trusts, including Oxford University Hospitals.
The study found the technology could be safely introduced into routine NHS pathology workflows.
It also reduced the need for some additional laboratory tests and, at some centres, meant patients could receive results around a day sooner.
Professor Clare Verrill, Professor of Cellular Pathology and Artificial Intelligence at the University of Oxford and Honorary Consultant in Cellular Pathology at Oxford University Hospitals, led the research.
She said: "Sometimes prostate cancer is difficult to spot. It can be a very small area, less than half a millimetre, and sometimes cancer can look quite similar to non-cancerous tissue.
"The AI works alongside histopathologists to reduce the chances of overlooking those areas, making more accurate diagnoses and helping to standardise assessments."
Rather than simply testing the software in isolation, researchers assessed how it performed during day-to-day NHS practice.
Professor Verrill said: "This study is one of the first examples of real-time use of AI in NHS histopathology workflows.
"You can have all the retrospective standalone performance studies in the world, but to actually get your hands on the technology and road-test it in the NHS, which is a complex environment, you won't know how it's going to perform."
Researchers found the AI accurately identified and graded prostate cancer and, in a small number of cases, changed a patient's diagnosis or treatment options.
The system is now being used routinely in Oxford's pathology laboratory.
Professor Verrill said: "The next step is this becoming part of everyday life for histopathology. More and more labs will take up this technology and use it and see the benefits that we outline in the study."