Pinpointing the heart’s trouble spots with artificial intelligence

Ventricular tachycardia is a heart disease in which the rhythm of the heart’s chambers is suddenly disrupted. Today, the disease is often treated with ablation – passing a thin tube through the blood vessels up to the heart and using heat-waves to remove the cells that cause rhythm disturbances.

Although ablation is the most effective treatment today, rhythm disturbances often return. Up to half of the people with ventricular tachycardia relapse within one year because not all of the offending cells have been removed.

AI can help doctors target the right cells – and reduce the risk of relapse.

More precisely, AI analyses the electrical signals in the heart and tells doctors exactly which cells are interfering with the rhythm – so the treatment hits the right cells right away.

“If we can use an algorithm to ensure that this happens, it will be a major advance in treating these people,” explains a researcher behind the study, Michele Orini, Senior Researcher from King’s College London, United Kingdom.

The research was a great multidisciplinary effort, bringing together health data scientists led by Dr Orini (King's College London), cardiologists led by Prof Lambiase (University College London) and Prof Tfelt-Hansen (University of Copenhagen), experts in cardiac physiology led by Prof Jespersen (University of Copenhagen), and cutting-edge facilities at Rigshospitalet, Copenhagen.

Ablation works – but not always on target

Before ablation, doctors make an electrical map of the heart using small measuring devices in catheters threaded through blood vessels to the heart.

Doctors view the electrical activity on a monitor to identify which cells are triggering the abnormal impulses.

Then the doctors can destroy these cells with, for example, radio waves. This method is often successful and the person is cured but is ineffective in other cases.

“Ablation is a life-saving treatment for many people, but it is also a complex treatment because it is not easy to identify the specific cells that are causing problems,” says Michele Orini.

Cracking the code of heart rhythm failure

The goal of Michele Orini and colleagues was to test whether an AI algorithm could support cardiologists in identifying the exact cells to destroy.

The researchers induced arrhythmia in 13 pigs and treated them as real patients – to determine whether the algorithm could pinpoint the diseased areas.

One advantage of using pigs is that researchers can collect data for much longer than would be possible with a human on the operating table.

Pigs’ hearts are similar in size and function to human hearts, making them well suited for such studies.

The researchers collected tens of thousands of electrical signals from each pig and used a machine-learning algorithm (random forest) to find patterns in the signals and identify the locations in the heart where ablation was likely to work best.

The trial showed that the algorithm could accurately guide doctors to the cells causing the heart rhythm disturbances.

Of four algorithms tested, the random forest model performed best.

Michele Orini calls the results promising.

“We imagine doctors working as they usually do – but with an algorithm in the background that analyses the signals and suggests exactly where to treat for the greatest effect,” he says.

From pigs’ hearts to patient care

Michele Orini says that the researchers have much work to do before the algorithm will play a role in treatment of people with ventricular tachycardia.

First, the researchers would like to further develop the model so that it can work with less complex data than is currently required.

In addition, the researchers would like to validate the use of the model in human trials.

Human trials are already underway – and preliminary results suggest that the algorithm works even better for humans than for animals.

“The goal is to be able to cure more people with ventricular tachycardia than we do today. However, some patients have disease that is so advanced and complex that they probably cannot be cured. The algorithm can also help doctors to distinguish between those who have a good chance of being cured – and those for whom treatment is unlikely to work,” concludes Michele Orini.