Potential for AI in NHS is ‘limitless’, expert says

The potential for the use of artificial intelligence (AI) in the NHS could be “limitless” but the pace of its rollout should “balance safety with efficiency”, according to an expert.

Dr Adam Dubis, associate professor and programme lead in digital health and entrepreneurship at University College London’s (UCL’s) Global Business School for Health, told the PA news agency AI could help in several areas, such as how long it takes until patients are seen.

He said: “The potential impact on the NHS is limitless. We are well aware of the long wait times to get into the GP and routine follow-up care.

“Having AI-enabled assistants will help triage these and provide necessary support for those waiting.

“A system optimised by AI which sees the sickest first, for whom intervention is still possible, is more time and cost efficient.”

In some parts of the NHS, AI is currently being used to analyse images, such as mammograms, and support radiologists in making decisions.

However, Dr Dubis said AI does have “several limitations”.

He added: “The biggest issue is that AI technology learns from what it has previously seen.

“We frequently read about the hundreds of thousands or low millions of health records being aggregated to develop some of these AI tools.

“However, when we compare these numbers to the tens of billions of images available for AI tools like self-driving cars, we see the problem.

“The errors in self-driving cars make international news, so we know that even with all that data, these tools are not perfect.

“Given the regulations around data sharing, it is very difficult to aggregate enough data to learn all parameters of humanity and disease.

“We need to continually work together to understand the power of AI and how it can improve healthcare efficiency and quality, and work to aggregate the data needed to become better.”

The Government recently committed £21 million to rolling out AI more widely on the NHS, with trusts invited to apply for a portion of the funding.

The Department of Health and Social Care (DHSC) said the technology could help cut NHS waiting lists ahead of winter.

On Tuesday, Dr Raj Jena, an oncologist at Cambridge University Hospitals NHS Foundation Trust, said AI could be used to treat more cancers.

He led a research team which found that a system he helped design was freeing up radiologists’ time.

Dr Jena created Osairis using a £500,000 grant from NHS AI Lab and it is currently being used in the head and neck and prostate cancer departments at Addenbrooke’s Hospital in Cambridge.

The system helps doctors draw around healthy organs prior to a patient being given radiotherapy. The crucial step ensures healthy tissue is protected.

Dr Jena now hopes the technology can be used for the likes of breast and lung cancer, as well as brain tumours.

And, in June, NHS England chief executive Amanda Pritchard said more uses for AI in the NHS were “on the horizon”.

However, shadow health secretary Wes Streeting argued that the technology is not being deployed fast enough.

Dr Dubis said the rollout of AI in the NHS should “balance safety with efficiency”.

He added: “I agree that we need to move faster, but this needs to be done in a safe way.

“Any traditional drug or tool that is implemented into a clinical workflow needs to go through extensive clinical trials to show safety and efficacy, and to provide a rationale for why it works.

“We need to apply the same standards to any AI tools which are responsible for patient care. Above all, we need to address safety and explain the ability to move forward.”

The debate around how AI should be regulated going forward is ongoing and Dr Dubis called for people to be educated about the technology and its potential.

He said: “We need to work on both the development, but also the education.

“For years, futurists have said that people’s jobs will be replaced within a few years, yet the mass layoffs have not come.

“There are technological reasons for this, but also human factors. Some jobs, like medicine, need to be done by an emotional and engaged entity.

“We are a long way removed from providing these sorts of assets to a computer.

“These trade-offs will provide some push-back and place a check on advancement.

“As long as the same levels of scrutiny are applied to AI as any other clinical tool, the progress will be safe going forward. It was not long ago that medicinal chemistry made rapid advances in new therapeutic methods.

“At that time there was the fear over what these substances would do. In the meantime, there were some failures, but there were also great advancements in health and life expectancy.

“AI will have the same life course as medicinal chemistry.

“Growth will be regulated by checks and balances both from official channels such as regulatory bodies and human factors – the needs and desires of the people.”