Artificial intelligence is better at spotting prostate cancer than hospital doctors, a groundbreaking study found.
Developed by experts, the computer system was trained and then tested on more than 10,000 prostate MRI examinations on patients.
Using the AI resulted in half fewer false positives and slashed the number of clinically insignificant cancers by a fifth when compared to radiologists, the research revealed.
Doctors believe it could help reduce overdiagnosis and prevent unnecessary surgery in the most common cancer among men, hugely benefitting any future screening programme.
Researchers predict using AI to help read scans will be crucial in addressing the rising demand in medical imaging worldwide.
The computer system was trained and then tested on more than 10,000 prostate MRI examinations on patients
More than 52,000 men are diagnosed with prostate cancer every year on average in the UK, making it the most common cancer in men. Around 12,000 men die every year from the disease — the equivalent of one every 45 minutes
However, as yet there is limited scientific evidence on whether it really works, which is stopping the widescale adoption of AI systems for prostate cancer diagnosis.
They wanted to test if state-of-the-art AI models, trained using thousands of patient examinations, were as good as radiologists when detecting clinically significant prostate cancer using MRI.
In the first study of its kind, an international team developed the technology using 10,207 MRI examinations from 9,129 patients in the Netherlands.
This was then tested on a further 1,000 patient scans to determine if the men had cancer and if so, how aggressive it was likely to be.
The technology was pitted against 62 radiologists from 20 countries, each with an average of five to ten years’ experience in interpreting prostate MRI scans.
The diagnoses were then checked against results using histopathology – analysis of cells under the microscope – and how the patients had fared an average of four to six years later.
It was found to be just as effective as radiologists at picking up more severe prostate cancers and gave half the number of false positives, which can lead to unnecessary biopsies.
AI also picked up 20 per cent fewer cases where the cancer was so low level, it was unlikely to trouble the patient in his lifetime, according to the paper published in The Lancet Oncology.
The global researchers, including from the Radboud University Medical Centre, Netherlands, conclude: ‘AI systems, when adequately trained and validated for a target population with thousands of patient cases, could potentially support the diagnostic pathway of prostate cancer management.
‘A clinical trial is required to determine if such a system translates to improvements in workflow efficiency, health-care equity, and patient outcomes.’
It comes as a landmark prostate cancer screening trial is getting underway in the UK, with a national screening programme likely to follow.
Transform will test the effectiveness of diagnosis techniques including MRIs when compared to the current standard, prostate specific antigen (PSA) blood test.
Prostate cancer is the most common cancer among men and second most deadly, responsible for around 12,000 deaths a year.
The sooner it is diagnosed, the better the outcomes with the majority of men alive five years later when diagnosed at its earliest state, compared with just half at stage four, when it has spread elsewhere.
Late diagnosis is also extremely costly for the NHS, with hormone treatments for advanced prostate cancer estimated to cost tens of millions each year.
Last year, the European Council agreed a recommendation that countries should consider the feasibility and effectiveness of prostate cancer screening based on PSA testing followed by an MRI scan.
Prostate Cancer UK is now calling for GPs to actively target men who are at increased risk – such as those with a history of cancer in the family or black men – to see if they want a PSA blood test.
The charity wants other trained healthcare professionals, such as practice nurses, to be trained in counselling men on the disease and testing options.
Chiara De Biase, Director of Health Services, Equity & Improvement at Prostate Cancer UK, said: ‘We know that a lot of men don’t come forward because they think they’ll be invited as part of routine tests – when this simply isn’t true.
‘The long-term answer is a screening programme for all men, and the results of our TRANSFORM trial will tell us the most effective way to do that.
‘Until then, we all need to work together to make sure men know their risk and what they can do about it.’