How AI could save the broken NHS

Telegraph

4 hours ago

Illustrations by Andreion De Castro
You might be frustrated by your recent run-in with a customer service bot but when it comes to healthcare, there's good reason to suggest that a wave of new AI tools could represent hope for our creaking health service, slashing backlogs and catching diseases years or even decades earlier, lengthening lives in the process.
This doesn't mean that your GP or consultant will soon be replaced by a silicon droid; the human element of healthcare will still be very much present. But, behind the scenes, AI is slowly beginning to reveal its prowess.
Last month, Health Secretary Wes Streeting announced that a new AI-powered blood test to detect cancer will be trialled on the NHS. Called miONCO-Dx, it can detect early warning signs of 12 of the most lethal cancers through identifying worrying traces of tiny molecules called microRNAs, which can change the activity of genes.
Proponents of AI's uses in healthcare often describe these tools as either 'a second set of eyes' or 'augmenting what a human doctor can do,' with Colin Rees, Professor of Gastroenterology at Newcastle University saying that he's enormously optimistic about their potential to limit preventable deaths.
'I think AI is going to transform healthcare like almost nothing else ever,' he says. 'Just in diagnostics, it's able to spot new patterns, look at blood tests and scans with a fresh perspective and, of course, it doesn't get tired or make mistakes.'
Already, AI is beginning to make a difference in disease-prevention for a broad range of conditions, from early detection of osteoporosis to helping overstretched radiologists prioritise which patients to examine first.
Let's take a closer look at what's happening right now, how technology may affect different diseases and common conditions, and what's likely to come later down the line.
Lungs
Since early 2022, a growing number of NHS Trusts across the UK have been using tools developed by a company called Annalise.ai to accelerate analysis of the many millions of chest X-rays taken every year.
With the UK currently suffering from a chronic shortage of radiologists – the latest data suggests that the NHS has 30 per cent fewer radiologists than it requires – AI has been trying to pick up the slack. In particular, according to Annalise.ai's CEO and co-founder Dr Aengus Tran, one of the goals has been to slash the time taken to diagnose cases of lung cancer – one of the most common cancers in the UK.
'The NHS has procured our AI across 131 hospitals to essentially run our system over chest X-rays, to see if the AI can find signs of cancer; perhaps subtle signs which might have been missed,' says Dr Tran. 'And if that's the case, those patients will be escalated to the very top of the list.'
The early signs are promising. Data collected from NHS Grampian, in Scotland, suggests that the tool has increased the proportion of lung cancers identified at earlier stages by nearly a third, and the technology is now taking on an even wider remit.
Because Annalise.ai's tools are capable of diagnosing more than 130 different lung-related conditions, from rib fractures to pneumonia to an enlarged heart, some NHS hospitals are already using them to work through their backlog of uninterpreted scans and help prioritise the most urgent cases. 'It means that patients with life-threatening illness and critical conditions are being seen sooner,' says Dr Tran. 'So, for example, in [hospitals in] Manchester, we're seeing that cases of pneumothorax [a collapsed lung] get flagged within less than a day, whereas previously that would have taken five or more days to get to.'
Heart
More than 1.5 million people in the UK have atrial fibrillation, a condition characterised by sporadic bursts of an irregular heartbeat. This can have serious long-term consequences, as blood can begin to collect within the heart, forming clots which can then travel to the brain. As a result, people with atrial fibrillation are also more at risk of suffering a stroke.
The problem, according to Rameen Shakur, a clinical cardiologist and Professor of Genomics & Precision Cardiovascular Medicine at the University of Brighton, is that many of these cases go undetected (an estimated 270,000 people in the UK have undiagnosed atrial fibrillation) because patients do not always present with symptoms when being examined in hospital.
To address this, Prof Shakur created a company called Cambridge Heartwear, which has designed a wearable monitor called Heartsense that records patient's heart rhythms to medical grade standard and checks for signs of atrial fibrillation using inbuilt AI algorithms. The device has since been tested in a clinical trial at a NHS hospital in Worthing, and Prof Shakur is hopeful that it can help to prevent some of the 100,000 strokes that occur each year in the UK alone.
'Once you find out that someone has atrial fibrillation, they can be prescribed anticoagulants [blood thinners] to reduce stroke risk, but that treatment is wholly dependent on catching it and getting a diagnosis,' he says. 'The advantage of this device is that it's discrete and malleable enough that you don't feel it. You can wear it while you're having a shower, when you're sleeping, when you're running to catch a bus – all times when an episode of atrial fibrillation might occur.'
Gut
Earlier this year, the NHS began offering biennial bowel screening to all people aged between 50 and 74, with the aim of detecting and preventing more cases of bowel cancer. According to Prof Rees, the health service now carries out around 700,000 colonoscopy examinations – where a person's bowel is examined using a small camera – every year.
As well as scanning for tumours, one of the aims of this procedure is to identify tiny precancerous warts called polyps, and remove them before they develop into anything more sinister. However, spotting polyps is not always straightforward. 'It requires a lot of concentration,' says Prof Rees. 'The bowel has got lots of folds in it, the large bowel is about a metre long and it has lots of blind spots. And there's very good evidence that missing these things leads to more patients developing cancer.'
To address this issue, Prof Rees recently led a pioneering trial in which gastroenterologists carrying out colonoscopies were given the additional assistance of an AI tool called GI Genius, which had been trained to spot abnormalities within the bowel using powerful neural networks. 'It scans the image which the specialist is looking at on a screen through their scope,' says Prof Rees. 'If it spots an abnormality, it flags it and puts a green box around it. I think of it as being a little bit like spellcheck in Microsoft Word.'
The trial, which involved more than 2,000 patients across 10 NHS Trusts, was successful: 30 per cent more polyps were detected, particularly a form of polyp which is quite notoriously difficult to spot via the naked eye alone. 'We call them sessile serrated polyps,' says Prof Rees. 'They're smaller, flatter polyps but they account for 30 per cent of bowel cancers, so they're particularly dangerous, and we found a significant increase in the detection of these bad guys.'
Bones
For many people in their 50s, a twinge of back pain might be dismissed as a sign of encroaching old age, but, in some cases, it could be a spinal fracture, which represents an early warning sign of degenerative bone disease.
When detected early, osteoporosis – a condition which causes the bones to become weaker and more fragile – is highly treatable. Now a new collaboration between academics, doctors, Royal Osteoporosis Society and the company Nanox.ai is investigating whether the company's technology can pinpoint potentially tell-tale fracture signs.
It could mean that, in future, whenever people in their 50s or 60s receive a CT scan, it will automatically be fed through AI technology to check for hints of osteoporosis.
Eyes
Can Google help prevent one of the most common causes of blindness? An ongoing partnership between Google Health, AI company DeepMind and Moorfields Eye Hospital is attempting to achieve just that through an AI system that aims to predict patients at particularly high risk of age-related macular degeneration – a condition caused by a progressive loss of cells in the retina – with the aim of identifying candidates for preventative treatment.
Because the eye acts as a window to the brain, other AI tools are being developed with the aim of detecting early signs of brain tumours or neurological diseases like Parkinson's, multiple sclerosis and Alzheimer's. According to Dr Siegfried Wagner of the UCL Institute of Ophthalmology and Moorfields Eye Hospital, AI can detect signs of multiple sclerosis through an asymmetry which only becomes apparent when scanning the left and right eyes. He predicts that in the future, such patients could be immediately referred for an MRI scan to investigate further.
'One eye might look completely normal, but AI can pick up a subtle thinning or thickening in the retinal nerve fibres of the other eye,' says Dr Wagner. 'It happens because in multiple sclerosis you get multiple small lesions in the brain, and they're more in one half of the brain than the other. So, you're seeing a reflection of what's happening in the brain, but in the eye.'
Brain
One of the biggest hopes for AI is that it could find a way to predict early warning signs for dementia many years before the onset of symptoms. Indeed, Will Whiteley, Professor of Neurology and Epidemiology at the University of Edinburgh, and a consultant neurologist in NHS Lothian, is leading a major new project to try to train an AI model to do just that, using a bank of CT and MRI scans collected through patients in Scotland over the course of 10 years.
'Right now, neurologists look for shrinkage in parts of the brain called the temporal lob, which are particularly responsible for memory, and a sign of Alzheimer's disease,' says Prof Whiteley. ' Vascular dementia affects the small vessels of the brain, and you can see that on a brain scan, too. But I hope that with these very powerful and flexible AI models, they'll be doing more, finding new things that we hadn't thought of as humans. We're not there yet, but wouldn't it be great if in future all 50-year-olds could take a test to tell you about your risk of dementia and whether you should or shouldn't take a medicine to prevent you getting it in later life?'