Scientists from the MRC London Institute of Medical Sciences (LMS) have used AI to identify genetic and environmental factors which cause the heart to stiffen, a common precursor to cardiovascular disease and heart failure. 

The study, led by the Computational Cardiac Imaging Group at the LMS, in collaboration with Bayer and the British Heart Foundation, used machine learning and artificial intelligence (AI) to analyse heart scans from thousands of people. 

They found that a range of factors, including several genetic and physical traits – such as ageing and diabetes – contribute significantly to the chambers of the heart becoming more stiff over time, making it less able to pump blood efficiently. 

The team say their findings, which are published in the journal Nature Cardiovascular Research, could ultimately help to spot people at increased risk of heart failure and potentially provide new targets for preventative medicines to treat patients earlier. 

Professor Declan O’Regan, who led the work, said: “Increased heart stiffness affects 50% of older adults causing reduced quality of life but there are few treatments available. AI offers a powerful new approach for analyzing heart scans and linking that information with genetics. These findings could accelerate the search for new treatments that target this early stage of heart failure”. 

In the study, the researchers analysed heart motion imaging from almost 40,000 participants of the UK Biobank – a huge database of biomedical information used for public health research. 

The team used deep learning to analyse MRI scans of beating hearts to determine signs of stiffness in the heart muscle. They cross referenced these with genetic data in what is known as a genome wide association study (GWAS). This enabled them to identify 9 locations in the genome linked to heart stiffness and determine molecular pathways that could be targeted therapeutically.  

As well as identifying genetic factors, the team cross referenced other datasets to determine other factors associated with heart stiffness. They found that older age, higher pulse rate, being of male sex and having diabetes are all significantly associated. 

Heart muscle stiffness is a common precursor to cardiac disease and heart failure that impacts the heart’s ability to refill with blood each time it beats, meaning oxygen is less efficiently transported around the body. “Heart muscle stiffness is poorly understood so efforts to uncover the molecular mechanisms behind it could enable the development of innovative therapies for many cardiovascular diseases,” says Declan. 

Deep learning heart movement analysis

The Computational Cardiac Imaging group were able to quantify the movements of the heart which meant the images could be read by a deep learning algorithm which they developed. “We are a very diverse team when it comes to areas of expertise. We have mathematicians, statisticians and computer scientists all working alongside geneticists, cell biologists and radiologists. It’s this exciting combination of disciplines that makes discoveries like this possible,” says Declan.

Which genes are associated with heart muscle stiffness? 

The group identified nine genes associated with heart muscle stiffness. These include ones associated with the mechanics of heart muscle cell contraction and relaxation as well as those associated with how the heart responds to mechanical stress. One gene that was identified encodes a protein that is known to play a key role in regulating heart function and has now also been highlighted as important in cardiac relaxation. This could be a potential therapeutic target, the authors of the study say. 

Read more

This study was published in Nature Cardiovascular Research.