Editorial – Unlocking the potential of artificial intelligence

After taking part in the EIT Health Think Tank Round Table Meeting in Ireland, Dr Pepijn van de Ven, Senior Lecturer in the Department of Electronic & Computer Engineering at the University of Limerick (UL) and Course Director for UL’s industry-driven online MSc in Artificial Intelligence, discussed how training and education will help unlock the potential of artificial intelligence (AI) in healthcare.

According to Dr Pepijn van de Ven, the use of AI in healthcare can generate huge efficiencies in terms of administration time, and support clinical decision-making. However, he notes, sometimes even the word AI sparks concern, and there is widespread misunderstanding about how the technology is – or might be – used in clinical practice. Here Dr van de Ven highlights that a re-education on the terminology of AI and its inclusion within medical curricula, will help change perceptions and ease acceptance and adoption of the technology across healthcare settings.

What is the biggest barrier to adoption of AI in healthcare settings?

Maybe the biggest barrier to adoption is the term AI itself. There is the perception that AI is all-seeing and capable of making decisions autonomously. There is a fear it will ‘take over’ because it conjures images of robots taking over the world. When you say AI, people don’t really understand what that means or what AI can (and can’t) do. What we’re talking about these days is really ‘machine learning’ or ‘learning from data’. And I think those two terms reflect much better what we’re really doing – we are learning from the data that’s available to us.

Listen to Dr Pepijn van de Ven describe in more detail the difference between AI and machine learning, and why continuing to educate people on this difference is important

Why is clinical education important in the adoption of AI within healthcare, and how is training best delivered?

If you’re introducing a new tool, then you need to make sure clinicians understand what the tool is and what these types of machines and tools can bring. It could be as simple as: what types of machine learning are there, what can they do, and what can’t they do?

This could be delivered along two parallel paths depending on where the individual is in their medical career. On the one hand, during medical training, an introductory module could be presented in the first year and then use cases and a deeper look at the ethics in the third year. On the other, for healthcare professionals (HCPs) who have already gone through education, short courses would allow them to upskill as part of ongoing continued professional development (CPD).

Finally, interactions between academia and hospitals tends to be on a consultant clinician level, which I think misses many other important stakeholders. Of course, consultants have insight into what can help them in their job, but I think a lot of gains of AI can be made at the nursing level. I see nurses as key, because they see what happens on the ground and have good insight into the processes being used in a hospital. And so, it’s tremendously important that they understand what AI can do and what it can’t. I would recommend that education on how AI can support medical practice should be mandatory for all clinical staff.

What other ways can HCPs be involved in ensuring widespread adoption of AI in healthcare?

Generally, in research studies, AI algorithms achieve very high accuracy; like 90% or higher. Despite this, they often don’t make it to the stage where they are widely used in clinical practice, possibly because we’re not thinking about the actual implementation of these types of technology soon enough. We need to make sure we understand how HCPs do their job, and the existing processes within which they work, as well as factors such as the funding model and data infrastructures. This means getting clinicians involved early in the process of developing AI tools to improve the chances of their success in real scenarios.

I also think identifying one or two small AI projects and making them happen in hospitals could aid adoption. Successful implementation of small-scale projects would likely be scaled up because when people see the power of a certain tool, they’ll start adapting to it quite naturally.

Is the desired change best driven or relevant at the national (member state) or EU level?

At the EU level, one obvious responsibility is funding to facilitate both research and the international collaboration of experts. Another thing that I think is really important at an EU level is the data structures, because wider use of AI is going to depend on loads of data being available to train algorithms. And I think we would want to achieve algorithms that are ideally useful anywhere in the world, but certainly across Europe.

Related to this, I think, is policy, and the ethics around the use of machine learning algorithms and personal information – it’s important that this is also driven at EU level. However, individual projects that will demonstrate how machine learning can be used in a clinical setting and how it will make a difference to medical practice, will be driven at a national level.

Along with other experts worldwide, Dr van de Ven is working to drive forward the adoption of AI in healthcare for the benefit of healthcare systems, HCPs, and patients alike. Despite the barriers yet to be overcome, he is certain that AI will have a big impact on the future of healthcare. Join the conversation and stay up to date with the latest thinking in AI on our Twitter and Facebook channels, @EITHealth, using the hashtag #EITHealthAI.