A physically fit 35-year-old, Sean Clifford had no reason to believe he was at risk of a stroke – or any other serious medical problem.
However, in 2023, keen to learn more about his health, the New York father-of-two decided to pay £2,500 to undergo a full-body MRI.
The scan, often referred to as a health MOT, is supposed to check the body for the early signs of disease. They rely on artificial intelligence (AI) software to spot subtle changes that might suggest something is not right.
And Sean believed he was in safe hands. The company behind the scan, called Prenuvo, has an impressive selection of celebrity supporters, including reality TV star Kim Kardashian, supermodel Cindy Crawford and actress Gwyneth Paltrow.
Moreover, the results were good. Prenuvo concluded that he showed no obvious signs of disease.
However, just eight months later, Sean suffered a catastrophic stroke that left him partially paralysed and brain damaged.
In September 2024, his family filed a lawsuit against Prenuvo after a radiologist contracted to reassess Sean’s scan found visible signs that arteries in his brain were narrowed – markedly raising his risk of a stroke in the near future. According to the lawsuit, which is still ongoing, had Prenuvo’s AI software flagged these narrowed arteries, Sean’s stroke could have been avoided.
Now experts say that thousands of NHS patients are at risk of similar life-threatening misdiagnoses.
In recent years the NHS has invested millions into AI scanning technology in an effort to speed up diagnoses. Ministers, who have described the technology as ‘game-changing’, argue that since AI can read scans quicker than doctors the move could slash NHS waiting lists.
Last year, it was revealed that AI was now being used in every NHS stroke unit in England to analyse and interpret brain scans, as well as half of all hospitals to help diagnose lung cancer.
But an expert has told The Mail on Sunday that growing evidence shows AI can miss the early signs of disease, leading to tragic misdiagnoses like Sean’s.
‘AI is a powerful tool, but it is not a reliable one,’ says Dr Joshua Henderson, a New Jersey-based psychologist and founder of the tech company Evidify, which analyses the impact AI has on the decision-making of radiographers.
‘The research consistently shows that these systems fail in ways that are unpredictable.’
The need for drastic measures to speed up NHS MRI scans is clear. The scans, which use powerful magnetic and radio waves to create detailed images of organs and tissues, are crucial for spotting the early signs of serious diseases including cancer, heart problems, stroke and fractures. Nearly five million are carried out every month on the NHS – however, there are currently major backlogs.
Patients are supposed to get results back within six weeks. However, last year, data revealed that nearly 400,000 patients at any one time have been waiting for their results for more than this period. Studies show that, for every month a cancer patient waits for MRI results, their risk of dying from cancer increases by about 10 per cent.
Research suggests this backlog is largely due to a severe shortage of radiologists – specialists trained to analyse scans. According to the Royal College of Radiologists, there are around 3,000 vacancies in the NHS – roughly a 30 per cent shortfall. Experts say this has led to the Government to embrace AI radiology software.
In theory, AI should not be used on its own to diagnose patients. Instead, radiologists are supposed to check the results to make sure the AI has made the correct call.
In one study, published in the medical journal Insights Into Imaging, AI was able to detect signs of stroke on an MRI scan in around 93 per cent of cases. This means it misses around one in 14 strokes. And, crucially, according to a 2024 study in the journal Radiology, specialists were only able to spot when AI had made the wrong decision in around a quarter of cases.
Dr Henderson argues that in the UK, where the tech is being rapidly adopted across the NHS, these missed mistakes could put lives at risk. He says: ‘When a screening result has been shaped by AI, patients deserve to know that a doctor exercised independent clinical judgment and did not simply defer to what the algorithm said.’
A spokesman for Prenuvo said: ‘We take any allegation seriously and are committed to addressing it through the legal process.’
A Department of Health and Social Care spokesperson said: ‘AI tools are used to assist – not replace – clinical decision-making, and all technologies deployed in the NHS must meet robust safety, effectiveness and regulatory standards before they are introduced.’
