Next generation diagnosis of coronary heart disease using ‘deep learning’ - Heart Research UK

Next generation diagnosis of coronary heart disease using ‘deep learning’

 

Novel and Emerging Technologies (NET) Grant – Dr Jack Lee, King’s College London

 

Amount: £238,762

 

Summary

Coronary heart disease (CHD) is where the coronary arteries that supply the heart muscle with blood become narrowed by a gradual build-up of fatty material. This can lead to angina and heart attacks, and is the leading cause of death in the UK. When a patient is admitted to a catheter lab for treatment for CHD, doctors must decide whether the artery should be re-opened physically with a stent, or in less severe cases, treated with medication.

There is much evidence that measuring the pressure drop across the coronary artery narrowing is a highly accurate way of deciding the best treatment. The test involves inserting a wire into the coronary artery which has a sensor to measure pressure. However, the majority of catheter labs in the UK do not currently measure pressure routinely. The reasons for this include risks to the patient, and extra time and cost of the procedure. The aim of this project is to make the pressure-based assessment of coronary artery narrowings safer, quicker and easier, using advanced computing processes.

Coronary angiography is the conventional method for looking at the coronary arteries and involves taking x-ray images of the blood vessels. This information can be combined with a computer model of blood flow to estimate the pressure drop, without carrying out invasive measurements on patients. There are already accurate methods to simulate the blood flow through blood vessels but they are time-consuming and require special training to perform.

In an alternative approach, this project will use an advanced computing algorithm known as ‘deep learning’. This is a type of artificial intelligence technique which will identify patterns from blood flow simulations in thousands of coronary arteries, so the computer ‘learns’ how the geometry of the narrowings affects the pressure pattern. In turn, this information may allow the pressure drop across the coronary artery narrowing to be calculated directly and in real-time from the angiography images. The team will then test the new method on real patient data to demonstrate its clinical usefulness.

The successful outcome of this research may help doctors decide on the best treatment for CHD using a test with reduced risk and less discomfort for patients. A fast and automatic method may also lead to shorter waiting times and cost savings for the NHS.