Research

Could artificial intelligence help treat heart failure?

Harnessing cutting-edge technologies could bring new hope for people with heart and circulatory disease. In the first of a new series, Sarah Brealey learns how artificial intelligence could help treat heart failure.

Artificial intelligence 3D heart

If you’re diagnosed with a disease, you probably want to know what the future holds. But with some conditions, it’s hard for your doctor to be certain.

Pulmonary hypertension is a rare condition that damages the right side of the heart. There’s no cure and, without treatment, it can lead to death or the need for a heart and lung transplant. Dilated cardiomyopathy is more common and can cause heart failure. It’s usually irreversible and is the biggest cause of heart transplants in the UK. For both, it’s difficult to predict which patients will go downhill rapidly, making it harder to offer the best treatments.

Using artificial intelligence to make treatment predictions

We’re funding Dr Declan O’Regan and his colleagues at Imperial College London to see if artificial intelligence can make better predictions than doctors.

Artificial intelligence - specifically a branch of it called machine learning – is being used in medicine to help with diagnosis. Computers might, for example, be better at interpreting heart scans.

We are good at diagnosis, but not so good at prognosis. We want to use machine learning to provide the missing piece of the puzzle.

Dr O’Regan wants to take that one step further. “We are good at diagnosis, but not so good at prognosis,” he says. “Making predictions about treatment that people with pulmonary hypertension will need is very difficult. We want to use machine learning to provide the missing piece of the puzzle.”

Computers can be ‘trained’ to make these predictions. You do this by feeding the computer information from hundreds or thousands of patients, plus instructions (an algorithm) on how to use that information. In Dr O’Regan’s work, this information is heart scans, genetic and other test results, and how long each patient survived. The computer starts to work out which factors affected the patients’ outlook, so it can make predictions about other patients.

“We have trained the computer to recognise features of the heart, so when we give it scans it can analyse them,” Dr O’Regan says. “These scans are in exquisite detail and the computer may be able to spot differences that are beyond human perception. It can also combine information from many different tests to give as accurate a picture as possible.”

Next steps and patient involvement

After starting with pulmonary hypertension, Dr O’Regan and his team are now looking at dilated cardiomyopathy. “If these techniques work then they should apply to different types of heart disease,” he says. “At the moment it is predicting survival, but in future it could, for example, predict stroke or other cardiovascular events.”

What really matters to patients is finding the right information and the right treatment at the right time

The team tested the machine with data from previous patients. They compared the predictions with what actually happened, so they know that it works. The next step will be to test it with a different set of data. If that works, they can start a clinical trial of current patients, where the computer makes predictions and the team follows up to see if these were correct.

In 2017, Dr O’Regan and his colleagues met a group of heart patients who were enthusiastic about being involved in this project. “What really matters to them is finding the right information and the right treatment at the right time,” he says.

A patient having a heart scan

Support from the BHF and computing progress

This research would be “impossible” without BHF support, says Dr O’Regan. “It all started eight years ago, with a small seed grant for equipment to take 3D pictures of the heart. That set the scene for this research. We have had two BHF grants since.”

We are not talking about machines taking over. It is not replacing radiologists or cardiologists; it is adding value and precision to what they already do

Using artificial intelligence in research has been made possible by developments in computing, including ‘deep learning’. Now, graphics processing units, developed for computer games, can perform thousands of tasks at the same time. “You don’t need a huge computer the size of a room any more: it can fit into a normal desktop computer in the clinic and extract data from many sources,” Dr O’Regan says.

“We are not talking about machines taking over,” he adds. “I think this is one area where machines could really benefit patients. It is not replacing radiologists or cardiologists; it is adding value and precision to what they already do.”

What do the terms mean?

  • Algorithm - An instruction, usually written in computer language, for how to solve a problem.
  • Artificial intelligence - A broad term for machines that do things that traditionally only human minds could do, such as understanding human speech or playing chess.
  • Deep learning - A type of machine learning in which machines learn for themselves what are the important characteristics in sets of data, particularly visual data.
  • Graphics processing unit - A type of electronic circuit for image processing, used in games consoles, computers and mobile phones, able to analyse large amounts of data at the same time.
  • Machine learning - A branch of artificial intelligence that gives computers the ability to learn without being explicitly programmed. It is based on complex models.

Dr Declan O'ReganCV - Dr Declan O'Regan

  • Consultant Radiologist
  • Head of the Robert Steiner MRI Facility, MRC London Institute of Medical Sciences
  • Director of Imaging Research, Imperial College Healthcare NHS Trust

More useful information