Trending Articles
Trending Topics
Trending
Imagine if your doctor would look at you when asking questions about your health instead of staring at their computer as they input your answers. It’s an odd problem caused by technology that could soon be solved by it – and yes, it’s happening in the NHS.
Now, the NHS isn’t actually new to AI. It’s being used for everything from scheduling patients to scanning patient records to look for signs of stroke. Sometimes it’s computer vision to spot disease in scans, other times an algorithm to sift through data.
Getting it right isn’t always easy – that’s true of AI anywhere, and any tech in the NHS – but the best and the brightest of the health service are looking for ways to solve real problems, be it budget, bureaucracy, or keeping us meat bags alive.
Shakeel Ahmad, lead stroke clinician for Wales; David Lowe, an emergency consultant and clinical director for innovation at the University of Glasgow; and Maeike Kusters, Consultant Paediatric Immunology and Innovation Officer at Great Ormond Street took to the stage at the Turing Institute’s AI UK conference to reveal how to get AI into the NHS – in a useful, sensible way.
David Lowe has a difficult job: bringing together everyone in a medical setting – including academics and industry – to decide what AI makes sense to actually put into use in the real world, considering the cost and the clinical effectiveness.
“We need a really clear understanding of our decision making around AI – it’s likely not the panacea for all of our challenges but really targeted interventions such as around stroke care is a really good example of how a technology can be used to build confidence.”
If clinicians see AI working to improve stroke outcomes, they will be more willing to use it for pulmonary embolisms, and so on. But, they also need to have key basics in place: there’s no point having AI examine health records to predict illness when those records are on “bits of paper”, notes Kusters, who points out that some of her paper files for very sick patients end up weighing as much as the child themselves.
In Wales, Ahmad is working on a platform called the Brainomix 360. It’s a silly name for a serious set of AI-powered tools that help clinicians decide what treatment each patient needs – a clot-busting drug called thrombolysis, or to go in and remove it mechanically, known as a thrombectomy.
“When you’re having a stroke, you’re losing two million brain cells a minute,” Ahmad says.
Normally, a patient suffering a stroke would go for a scan and then wait for up to an hour for it to be analysed. The AI system looks for a blockage, and sends the images with a notification to the consultant to look at on their phone, allowing them to make a call on the treatment path immediately.
Anything that helps speed up diagnosis can help speed recovery. “For me it’s about enhancing the clinical pathway… so we can make a quick decision,” he says.
The results are impressive, with hospitals already topping targets for getting the drug to patients. “if you can treat them and treat them quickly and restore their function, that’s a huge gain,” he adds. “So not just the healthcare system, to society or more as well.”
Kusters treats children who have rare diseases at Great Ormond Street Hospital, but the aim of the innovation department at that famous institution is to develop ideas that could be applicable to all doctors.
One example is the ambient conversation tool. “What you notice if you come into hospital and have a clinical appointment, the doctor brings you into the room – ‘the doctor will see you now,’ but they are looking straight at their computer,” she notes.
They type, look up information, while listening to you from one side of their head, she says. “Is that how we speak to each other – it’s not,” she says.
From this complaint, the team built an ambient conversation tool to solve the problem and let physicians actually look directly at their patients. It uses a speech to text tool with two mics – she works with children and they don’t sit still, so two microphones were required to cover the room well.
Recording can start even before the patient comes in, with the physician working through their background material. When the patient arrives, the doctor can speak directly to them, move around the room to do an examination, and everything is captured and translated into the clinician’s own template.
“This tool collects medical data – you can talk about football, your favourite book, it doesn’t really matter,” she says. “It collects the medical data, translates it into this template and makes notes. You can then just copy the notes into your electronic patient record (EPR).”
The system was designed in house and sits outside the EPR. That’s for a good reason: plenty of commercial patient record systems actually have AI transcription tools, but they’re expensive. “It’s not something we can afford within the NHS,” she says. “So we need to build our own tools, homegrown UK stuff.”
