Future heart and stroke risk predicted by the touch of a button

By | December 14, 2019

Pinpointing your future risk of heart attack or stroke with a simple bone scan may soon be a reality, as researchers broach a new frontier for medical research.

The National Health and Medical Research Council will provide $ 467,980 to fund a three-year collaborative project led by a team of Edith Cowan University (ECU) researchers.

The team will investigate how machine learning can automate the task of identifying and quantifying an individual’s blood vessel disease from a routine scan, years before symptoms arise.

This project builds on the recent work led by ECU researcher Associate Professor Joshua Lewis that machines used for routine bone density scans can also be used to identify the presence of calcium in a major artery in the abdomen, known as abdominal aortic calcification (AAC). AAC is a strong predictor of cardiovascular disease.

Professor Lewis said the NHMRC project would provide a window to identify high-risk people and intervene years before they suffer the devastating consequences of heart attack or stroke.

“Given each year more than 350,000 older Australian women undertake bone density testing, adding this scan at the time of bone density testing may lead to a paradigm shift in the way we monitor and prevent clinical cardiovascular disease,” Professor Lewis said.

“Using machine learning to scale up our AAC test will be a game changer, moving us from a reliance on trained experts to manually process each scan to getting results at the touch of a button.”

Cheap, safe, quick and reproducible test

Professor David Suter from ECU’s School of Science will lead the development of the deep learning architectures needed to automate the image analysis.

“This will be the first fully automated tool to routinely image and identify this disease,” Professor Suter said.

“We are well positioned to develop an accurate visual recognition of AAC in the scans, with access to over 11,000 scans that have already been assessed and accurately labelled by world experts.”

Professor Suter said the visual processing of the scans was incredibly complex.

“Each scan is different and so the program will need to find visual cues to understand where to look for the required signs of vascular disease,” Professor Suter said.

“We hope to ultimately develop a cheap, safe, quick and reproducible test for visualising vascular calcification that will save lives.”

Edith Cowan University will collaborate with the Universities of Manchester, Western Australia, Southhampton, and Minnesota, INSERM and the Hinda and Arthur Marcus Institute for Aging Research, a research affiliate of Harvard Medical School on the project.