With an introduction from editor Helena Beer, consultant radiologist Dr Oliver Hulson discusses an advanced artificial intelligence and machine learning software system being piloted at Leeds Teaching Hospitals NHS Trust to help improve prostate cancer diagnosis by accurately detecting prostate cancer lesions from MRI scans.
The potential for artificial intelligence (AI) to streamline cancer pathways and reduce the waiting times for biopsy is well known, but real-world use cases for AI in cancer diagnosis remain limited at present – something that Dr Oliver Hulson, consultant radiologist at Leeds Teaching Hospitals NHS Trust, is working to change.
As time progresses, the evidence base for AI-supported cancer diagnosis is growing, particularly in lung cancer detection on chest x-ray and lung computed tomography (CT), breast cancer detection on mammography and prostate cancer detection in prostate magnetic resonance imaging (MRI).
Dr Hulson’s focus is on prostate cancer as these patients face a complex diagnostic pathway. His aim is to use the pilot Prostate Intelligence (Pi) AI tool to accurately identify prostate cancer on MRI scans and facilitate faster diagnosis, faster treatment and better outcomes for patients.
How might the use of AI in prostate cancer diagnosis lead to better patient outcomes?
I am particularly interested in the use of AI to facilitate a streamlined prostate cancer pathway and reduce some of the inherent friction seen in our workflows currently.
My hope is that AI tools such as Pi will potentially allow us to risk stratify and triage men, such that men with significant cancer will be highlighted first and potentially be offered a biopsy the same day.
A ‘one stop shop’ such as this would mean men could potentially have their MRI scan in the morning, it be reviewed by the radiologist with the benefit of the AI tool, and, if needed, plan for their biopsy that afternoon.
Performing all their investigations in a single day, rather than over weeks, as is the case currently, would, of course, reduce their anxiety whilst waiting for tests and would provide their results as quickly as possible.
We are looking to trial this innovative approach in Leeds this year.
How does the Pi AI tool’s performance compare to traditional radiologist assessments for detecting prostate cancer lesions?
The PAIR-1 trial, a large multicentre, retrospective study, showed that Pi carried a high sensitivity (95%) and moderate specificity (67%) for the detection of clinically significant prostate cancer.
These results are certainly comparable with the performance of a radiologist, and the high sensitivity is such that the likelihood of cancer being missed with the tool is very low.
The ongoing study will compare the AI-generated results from the Pi software against real-world outcomes for 100 patients who have recently completed the prostate cancer pathway.
The Pi software assists radiologists by highlighting potential areas of concern on MRI scans and assessing risk scores and prostate size, which can impact biopsy and treatment decisions.
What are the practical considerations for integrating the Pi tool into existing radiology workflows?
The main limitations with the implementation of these tools into our pathways relate to integration with our current IT systems, including our picture archiving and communication system software and MRI scanning software.
Healthcare AI companies are able to provide support with integration and implementation, although such assistance typically comes with increased associated costs, which must be taken into account.
Radiologists will also need to be trained in the use of the software, and how to effectively use the tool to improve their reports. Automation in prostate volume contouring, prostate specific antigen density calculation and region of interest detection can potentially improve efficiency and productivity.
How could Pi support or enhance collaboration between members of the multidisciplinary team (MDT) to improve prostate cancer diagnosis and care planning?
AI in prostate MRI can enhance multidisciplinary collaboration by providing a shared, standardised radiology report, highlighting the key findings and any regions of interest for biopsy, which can then be exported for utilisation in a fusion biopsy platform.
AI may also help to reduce variability and discordance across reporters. AI-augmented structured reports and visual tools can also improve communication in MDT meetings, enabling more efficient decision-making and supporting more confident, person-centred diagnosis and care planning.
Since the pilot's initiation in September 2024, what measurable outcomes or advancements have been observed in the integration and effectiveness of the Pi tool?
We launched our ‘one stop shop’ for men with suspected prostate cancer in March of this year, and it is hoped that if this proof of concept is successful, we can then look to scale and expand the service, and thereby meet the 28-day faster diagnostic standard and 62-day cancer treatment targets set by NHS England.
The feedback we have had from patients who have been involved in this pilot so far has been uniformly positive, and has given us the enthusiasm and motivation needed to drive this initiative forward for even more men.
The West Yorkshire and Harrogate Cancer Alliance and similar organisations across the country are very interested to see the initial results from our pilot to see if the approach can be adopted in other centres across the region and nationwide.
What steps are necessary to evaluate the tool’s scalability across different NHS Trusts and its applicability to diverse patient populations?
The main challenges relate to the clinical- and cost-effectiveness of such tools, and if AI can also be applied in the emerging setting of MRI screening for prostate cancer.
These tools have been trained on large datasets, but the question remains as to whether these data sets provide sufficient diversity to be applicable across a broad range of ethnicities and patient demographics.
What are the next steps in research and development to enhance the tool’s capabilities?
We hope that if we can prove the concept of an AI-facilitated same-day MRI and biopsy service, this will become the standard of care for all men across the UK, improving cancer detection and reducing delays in a service under significant pressure.