AI (artificial intelligence) in heart healthcare

What is AI (artificial intelligence) and how is it used in healthcare?

AI is when computers are used to do tasks that would normally require human intelligence.

In healthcare, one use of AI is to analyse large sets of information (data) to help diagnose conditions or predict someone’s risk of developing them.

AI is especially good at spotting subtle patterns in data that humans might miss, as AI models can look and learn from millions of points of information at once. You could call this ‘superhuman AI’. It’s these models that will be particularly powerful tools to improve diagnosis and treatment in future.

Is AI already being used in the NHS?

While many AI tools are still in development, some are already being used in clinical settings.

For example, tools that transcribe and summarise conversations between doctors and patients are helping to save time on administrative tasks like writing letters to GPs. All the information remains confidential and is checked for accuracy by the doctor.

But the more advanced applications – such as using AI to detect heart conditions or predict future health risks – are still being tested and are a few years away from routine use.

How can AI improve diagnosis in heart healthcare?

AI could help us to make earlier diagnoses – hopefully before people get really ill.

For example, my research group, funded by British Heart Foundation, has developed an AI model, known as AI-ECG risk estimation, or AIRE, to analyse ECGs (electrocardiograms). These are simple, widely used tests that record the heart’s electrical activity.

Traditionally ECGs are used to diagnose only a handful of conditions such as heart attacks and irregular heart rhythms (arrhythmias) which show up in obvious ways on the test.

But AI can pull far more information from the same test and pick up subtle signals that humans cannot.

The next step is figuring out how to integrate this into the NHS.

The AI model we’ve developed, trained on millions of ECGs from around the world, can now detect over a dozen different conditions, including heart failure, valve disease, high blood pressure, diabetes and kidney disease.

It means a simple, cheap test can now be used to pick up a wide range of issues, often earlier than we could before.

We’ve developed the technology and we’re just starting to test it in NHS hospitals to see how well it works in a real-world setting. The next step is figuring out how to integrate this into the NHS.

In what other areas in heart healthcare could AI be used?

AI can be applied to almost any type of health data, from GP records and blood tests to imaging scans and even genetics.

For example, some of my colleagues are using AI to analyse CT scans of the heart’s arteries and look for patterns that might show which arteries are mostly likely to become blocked in future, meaning action can be taken to prevent a heart attack before it happens.

Others are using AI to screen GP records to identify people who are at higher risk of atrial fibrillation and who might benefit from further testing.

Can AI help guide treatment too?

Yes, absolutely. As a doctor, I specialise in treating abnormal heart rhythms. To treat atrial fibrillation, for example, I currently do a procedure called catheter ablation, which destroys small areas of heart tissue thought to be causing the irregular rhythm.

Right now, we use a one-size-fits-all approach for atrial fibrillation ablations, which is only successful about 50 per cent of the time.

AI could help us personalise treatment by pinpointing exactly where in each individual’s heart to treat. That could make a big difference to how successful they are.

AI could also help decide who should receive treatments like implantable cardioverter defibrillators (ICDs), which help treat dangerously abnormal heart rhythms that can cause a cardiac arrest.

The challenge is knowing who’s at highest risk. AI could analyse ECGs and heart scans to help us make better decisions.

In the future, AI might even help us to discover new medicines by analysing genetic data and proteins to identify new treatment targets. That’s still a bit further off, but it’s an exciting area of research.

What about wearable technology and AI?

Wearables like smartwatches already record ECGs, and if these wearables can be made more reliable and clinically accurate in future, I think there’s huge potential in combining them with AI.

A one-off hospital ECG is just a snapshot, but wearables can provide continuous data.

And of course, AI is able to analyse that large quantity of data in ways that a human doctor cannot.

What are the concerns around using AI in healthcare?

AI will only be useful if patients as well as doctors accept its use. That’s why, in developing AI models, we set up a public and patient involvement group.

We speak to patients and relatives about our work to get opinions about how to conduct our studies and ideas of where and how AI use is acceptable.

In my experience, people are keen for anything that can improve their heartcare and people are keen to have earlier diagnoses and better risk prediction.

The way to address bias in AI models is to do the science properly

The way to address this is to actually do the science properly. That’s why we’ve made sure our datasets include people from different ethnic backgrounds and genders.

Will AI replace doctors?

AI will not replace doctors, but it will help us be better doctors. It can help us make earlier diagnoses, choose the right treatments, and save time on administrative tasks.

That means we can focus more on what really matters – caring for our patients.

What to read next...