The use of AI to identify prostate cancer is currently being validated in the first retrospective study of its kind in the UK.
Real-world data from hospitals and scanner types is being scrutinised, to show how the technology could improve accuracy of diagnosis, and potentially help over 100,000 patients per year in the UK alone.
Prostate cancer is the most common cancer in men in Europe, Africa and the Americas, with 1.4 million diagnosed worldwide each year and 375,000 deaths. In the UK, prostate cancer now kills more patients each year than breast cancer, and 56% of cases in England are detected late, when the disease is high-risk or advanced. Improved ways of screening for and diagnosing the disease are needed to help detect the disease earlier and offer patients better treatment and outcomes.
The study is a collaboration between Hampshire Hospitals NHS Foundation Trust (HHFT) and Lucida Medical Ltd, whose (Prostate Intelligence) technology uses radiogenomics, machine learning and image processing to analyse magnetic resonance imaging (MRI) scans. Results presented at the European Congress of Radiology (ECR) in March 2021 indicate that it can help automate labour-intensive tasks such as marking out lesions, and avoid unnecessary invasive biopsies, with unprecedented accuracy and consistency. The technology recently received the CE mark, allowing for use in hospitals.
For the NHS to adopt this technology, its accuracy needs to be proven in a range of settings. Most studies on AI for radiology to date have used data from single centres, but it is known that performance can vary between hospitals and with equipment from different scanner types. This collaborative research led by HHFT will specifically test how the software would work at a range of different NHS hospitals and with all major MRI scanner manufacturers.
The PAIR-1 (Prostate AI Research – 1) study will collect data on 2100 patients who were diagnosed at seven different centres. All data will be de-identified so that the researchers will not be able to link it to individuals, following a strict protocol approved by the UK Health Research Authority. This work is especially valuable because it will allow the software to be tested with the same types of patients, scanners and hospitals that would be encountered in practical clinical use, but without impacting on patients’ care.
Prof Richard Hindley, consultant urologist and lead for innovation at HHFT, said: “We are delighted that HHFT is able to work with Lucida Medical on this collaborative study. It will gather real-world information on the prostate cancer diagnostic pathway across a range of NHS Hospitals, and is a natural evolution for our trust having been leading participants in the PROMIS trial (2015-2017) which provided the evidence to support the incorporation of prostate MRI into the assessment of any man with suspected prostate cancer.”
Dr Aarti Shah, consultant radiologist at HHFT and chief investigator for the PAIR-1 study, added: “Reviewing prostate MRI requires experience and expertise to ensure that the right patients have a biopsy as well as to help target biopsies to maximise the chances of finding significant cancers. AI has exciting potential to enhance the processes of screening and treating patients, and this study will provide strong evidence on its performance.”
Lucida Medical co-founder and Chief Medical Officer Prof Evis Sala, Professor of Oncological Imaging at the University of Cambridge, said: “Now Pi has CE marking, clinical studies such as this are crucial to demonstrate the performance of the system in real-world clinical use. HHFT has brought together a group of Trusts representative of the wide range of settings across the NHS, from major teaching centres to district general hospitals.”