The data frontier: How AI is poised to improve NHS diagnostics and treatment times

In May 2023, Yvonne Cook was attending what she thought would be a routine mammogram appointment. In the waiting room, she noticed a sign informing patients that a project was underway involving artificial intelligence (AI) assisting in reviewing patients’ X-rays. 

Participation was optional, and Yvonne, who is in her 60s and from Aberdeen, says it didn’t occur to her to opt out. “I think it said that AI would be utilised as part of the research project to review the mammogram, and I just thought, why not?” 

Shortly after her appointment, she received a recall letter requesting additional imaging – the wording was intentionally gentle.  

“When I arrived for that appointment, they said that it was the AI part of the analysis that had picked up something. I had a scan, and the consultant confirmed that the AI diagnosis was correct, that there was a small, Grade 2 tumour there, too small to be detected by the human eye.”  

It was lobular breast cancer, the second most common type.  

“You would expect a negative and emotional reaction to, ‘Oh my goodness, I have cancer,’ but overwhelmingly I just felt incredibly lucky that I was part of the research programme and that it had been picked up at this early stage.  

“The fact that it started with me feeling incredibly lucky, I think, really helped to shape my positive approach to the treatment and everything that followed.” 

Having immediately been placed on oestrogen-blocking medication to inhibit the growth of the tumour, which doctors determined was aggressive, surgery took place in December 2023. It was followed up with a second operation to remove more breast tissue in January 2024 to achieve clearer margins. By May 2024, she had completed her treatment, and by then, it had become clear to her how crucial the AI screening had been.  

“Had the AI not picked up the small tumour when it did, then either it would have been discovered at my next routine mammogram three years later, or I would have picked it up when it had grown to a stage that I was able to feel it,” Yvonne says.  

“If that had been the scenario, then it’s likely that the surgery would have been more invasive. The cancer could have spread, it could have involved chemotherapy, and a much longer recovery time with more impact on my life.  

“As it was, beyond some initial joint and muscle pain I experienced as a result of the oestrogen blocker, the impact on me wasn’t as significant as I feared. I was able to keep working throughout all of my treatment.” 

It was the AI deep-learning tool, known as Mia (Mammography Intelligent Assessment), that spotted Yvonne’s cancer. Developed by Kheiron Medical Technologies, it is designed as a third eye for clinicians who examine the results of breast screening scans. It is trained on large datasets of mammography images, and the algorithm evaluates each scan and produces a risk score alongside visual markers highlighting areas that may require closer examination by radiologists. 

The trial, named GEMINI (Grampian’s Evaluation of Mia in an Innovative National Breast Screening Initiative), was run in collaboration by NHS Grampian and the University of Aberdeen between February and October 2023. One of the key aims was to draw attention to subtle abnormalities that might otherwise be difficult to detect, particularly when tumours are extremely small or present in complex tissue structures, like in Yvonne’s case. More than 10,000 mammograms from routine check-ups were analysed during the trial. 

Earlier this month, the UK’s first comprehensive study into the technology was published in Nature Cancer by a team of scientists, clinicians, and software developers from the University of Aberdeen, NHS Grampian, and Kherion Medical Technologies. It found the AI tool increased detection by over 10 per cent and has the potential to reduce the workload of healthcare workers by more than 30 per cent compared to the current clinical process.  

Another headline finding was that it reduced the time to notify affected women from 14 days to just three. Authors of the research have described this development as “hugely significant” and say it will enable earlier treatment, which has “a greater likelihood of treatment success”. 

Dr Clarisse de Vries, the study’s lead author, a lecturer in data science at the University of Glasgow and a former research fellow at the University of Aberdeen, says there were 11 instances that clinicians did not think women needed to be recalled that were flagged by Mia and led to cancer diagnoses that would have otherwise been missed.  

She explains that technologies like Mia will be used to increase detection, bring forward the beginning of treatment, and reduce the overall anxiety of patients. “In the general process of breast cancer screening, in a group of women about five per cent will be recalled for further investigation based on potential signs of malignancy, but one in five of them will be diagnosed with cancer,” de Vries says. 

“Those additional investigations could be another mammogram, ultrasound, or a biopsy. Ideally, you want to increase cancer detection, but not the number of women going under investigations, because we know that causes women stress and anxiety.” 

AI technologies like Mia are attracting growing attention within the NHS as cancer services come under increasing pressure. Across Scotland, demand for diagnosis and treatment continues to rise, with a record number of cancer cases recorded in 2023.

At the same time, the workforce responsible for delivering cancer care remains stretched. Over the past five years, health boards have consistently struggled to meet the national standard that 95 per cent of patients should begin treatment within 62 days of an urgent referral. 

De Vries says during the GEMINI trial, using data, analysts were able to model how the tool would be able to cut down the strain on the workforce in various configurations.  

“In one instance, we looked at replacing one of the human readers with AI to read the mammogram. But what’s neat is that we looked at 17 different scenarios, so if a clinical centre has specific challenges in an area, the AI can be tailored to suit its needs.” 

Professor Richard Simcock, chief medical officer at Macmillan Cancer Support, is excited about the potential for AI in the NHS, telling Holyrood its growth is “genuinely exponential” and because most people have seen it develop through use in their own lives “most people don’t struggle understanding that”.  

“The GEMINI study is now something that has been well tested. It’s now no longer about testing; it’s about rollout and implementation. 
“The questions are now not about safety; it’s about IT interoperability and system capacity.” 

He points out several areas where AI is being used in other countries and in the private sector to decrease the stress on the workforce and increase the human interaction time between a clinician and a patient. One of the easier wins that he sees would be the rollout of ambient scribing technology, a sort of AI-powered dictation software.

“There is an NHS-approved vendors list, which has at least ten options,” Simcock says. 

“These are all software that listen to the consultation, and at the end of it produce a letter, a summary, and in some cases will triage the tasks that result from it. In private practice, we are seeing them running with it. 

“However, I see a little bit of procurement paralysis, and I think that comes from a number of worries, one of them being people in IT, who are already overburdened and are often the smallest departments in the hospital. There is also an anxiety over when to put your flag down on a technology because we know it is moving so exponentially.” 

The key to the rollout of AI in clinical settings is access to good, reliable, linked data that can be used in trials. Simcock explains that’s why GEMINI has been one of the first major trials seen in Scotland, because there are over two million mammograms taken each year in the UK, which the AI needs to be able to train its algorithm.  

De Vries says that Scotland’s advantage is the data linkage. “For over 10 years, Scotland’s breast cancer screening data has been digital, whereas in England some of it is still paper-based. So for data scientists like me, we can much more easily look at the data and carry out statistical analysis.” 

Scotland’s advantage is that its health data is also in large part due to the infrastructure. Across four regions, there are data services continually collecting data from health and social care with the primary goal of providing researchers with “research-ready” data. It is important to have data from different regions of Scotland, as increasingly AI tools are considering environmental factors that could impact a population of people.  

DataLoch is the service for the south-east Scotland region, and its clinical lead, Atul Anand, says the “idea is we have created the environment to allow someone to do a study across the four regions”. 

He says: “In GEMINI’s case, they could apply the data that is harmonised from Edinburgh or Glasgow to provide external testing, and probe whether the technology works with a different population. 

“That’s what we really want to do – have something be built locally and have them able to access data from somewhere else in Scotland, that is well aligned and understood to the data they are using locally.” 

It’s clear that Scotland is emerging as a place where we are seeing a new frontier of health technology trained on reliable data. With a strong ecosystem of academia, data, and a willing population, the GEMINI study offers a glimpse into what the next 10 years in the NHS could look like, as well as what the potential solutions to well-documented healthcare challenges could be.  

For Yvonne, it has already had a huge impact on her life, and although she says she has faith in the current screening system, she says: “At the back of my mind when I go in for my checkups, I am thinking if there’s no AI as a second check, what if there’s something small there that isn’t picked up early?  

“There could be many other women who would be in the same situation as me. Being caught very early through AI and dealt with quickly and less invasively way is a huge bonus.  

“The flip side is that if the cancer isn’t detected, the treatment protocol could be quite different – not just the impact on my life, but also the cost to the NHS of treating a tumour that is much further advanced. 

“I just feel incredibly lucky. Lucky that AI was used, and lucky that it caught something so small at exactly the right time.”