Mapping arteries to treat heart attacks before they happen

Project Leader: Prof. Andrew Ooi, Dr Eric Poon, Dr Shuang Zhu, Prof. Peter Barlis and Dr Vikas Thondapu, University of Melbourne

According to the Australian Bureau of Statistics 2017–18 National Health Survey, an estimated 1.2 million (six per cent) of Australian adults aged 18 years and over had one or more conditions related to heart or vascular disease, including stroke.

University of Melbourne’s researchers Professor Andrew Ooi, Dr Eric Poon and Dr Shuang Zhu have partnered with clinicians Professor Peter Barlis and Dr Vikas Thondapu in pioneering virtual tools using mathematics and Pawsey supercomputing to develop a method to rapidly and accurately detect coronary artery disease and inform treatment well before fatal heart attacks occur.

Their approach is determining the extent of arterial blockages and inform the best course of treatment recommendations.

 
core hours allocated
 
journal articles
 
peer-reviewed conference papers
Partner Institution: University of Melbourne System: Magnus, Zeus Areas of science: Biology, Computational Fluid Dynamics

The Challenge

Coronary artery disease, or narrowing of the coronary arteries through plaque buildup, is the most common cause of death globally through causing heart attacks.  Many medical approaches to assess the degree of artery narrowing are invasive and provide limited information to guide treatment.

The current gold standard to assess arterial blockages involves catheter-based coronary angiography, which takes low-resolution 2D pictures of the artery.  This doesn’t give a functional assessment of blood flow though, which requires an even more invasive procedure to insert a pressure wire into the artery.

The wire measures the pressure at just two points, but the pressure drop between the points can be used to infer the size of the blockage and hence the volume of the blood flow getting through.  Neither technique gives the exact location or geometry of the blockage.

To improve on this, Professor Ooi is using information collected by interventional cardiologist Professor Barlis during angiography and combining it with optical coherence tomography (OCT) medical imaging data, the highest resolution 2D arterial cross-section imaging available, to build accurate 3D computer models of diseased arteries.  Computational fluid dynamics simulations can then be run to predict blood flow in the artery, giving a complete picture of flow restriction and pressure at every point, accurately mapping blockages in much greater detail to guide treatment.

The technology has already been demonstrated and compares extremely well against existing diagnostic methods.  The challenge, Professor Ooi explains: “is to provide the answers fast enough to be clinically useful for cardiologists, who often need to make decisions to operate on a critical patient within minutes, not days”.

The Solution

Reconstructing the artery inside a computer is the first part of the process. “It takes a few days on a local computer, but it’s not convenient to build the simulations much faster on the Pawsey supercomputers yet,” admits Professor Ooi. “Reconstruction currently requires a lot of manual work so it’s easier to do that locally.”

 

 

Pawsey supercomputing facilities are being used to run the computational fluid dynamic simulations on the complete artery reconstructions.  “We can get the full flow field, pressure at every point, turbulence, wall shear stress – all of the blood flow behaviour that is physiologically significant, created from the detailed shape of the artery.  But even using hundreds of processors at Pawsey, it’s taking too long to do the computations for the cardiologists to use it.  They want answers in minutes,” Says Professor Ooi.

“Right now, we’re working out the minimum information that they need – probably just a pressure map rather than a full flow field – and developing algorithms to predict only that, much faster.”

The Outcome

The project aims to develop an efficient computational tool to give cardiologists the information they need to diagnose and recommend treatment in a clinically useful timeframe.  With information available about artery shape and obstruction, and assessment of blood flow around the blockage, they can decide whether a patient can be treated with medication, or if expansion of the artery is required via the insertion of a stent.

“We’re combining skills in image processing, AI and computational fluid dynamics to work towards a new gold standard that will provide trained cardiologists with more detailed information to guide them in the diagnosis and treatment of coronary artery diseases”, Professor Ooi forecasts.  “If we can automate and simplify the process enough to enable fast, high-volume use in a clinical setting, we can avoid invasive pressure wire tests and reduce the risk of complications associated with diagnostic tests, while providing rapid and accurate detection of coronary artery disease and improved treatment well before fatal heart attacks occur.”

Prof. Andrew Ooi, Dr Eric Poon, Dr Shuang Zhu, Prof. Peter Barlis and Dr Vikas Thondapu, University of Melbourne,
Project Leader.