The eyes may be the window to the heart as well as the soul – particularly, to whether that heart is at risk of an infarct, researchers reported last week at the annual congress of the European Society of Human Genetics.
Combining imaging the blood vessels of the retina and genomic data with classical risk factors was able to more accurately classify the risk of coronary artery disease (CAD), and of heart attack, than those risk factors by themselves.
The team looked at the branching patterns of the blood vessels in the retina and calculated a score that captured overall branching complexity from those data. In their work, lower branching complexity was associated with higher risk of CAD and myocardial infarction (MI).
The same factors that damage the blood vessels in CAD affect the vasculature in the rest of the body. But imaging the coronary arteries themselves is complicated and expensive, and done only when there are already clear indications that there is a problem.
Imaging the blood vessels of the retina, on the other hand, is a routine part of annual eye exams, meaning that at least those individuals who get their eye exams have up-to-date vascular imaging available already.
"There have been multiple attempts to improve CAD and MI risk predictive models by accounting for retinal vascular traits, but these showed no significant improvement when compared with established models. In our case, we found that the clinical MI definition – the diagnostic codes that describe myocardial infarction events in medical records – is central to the successful development of predictive models, underpinning the need for developing robust disease definitions in large studies such as UKB. Once we validated our MI definition, we found that our model worked extremely well," said presenting author Ana Villaplana-Velasco, a graduate student at the University of Edinburgh's Usher and Roslin Institutes.
Speaking to the U.K.'s Science Media Center about the research, James Ware, Cardiologist, Reader in Genomic Medicine at Imperial College London and MRC Investigator, MRC London Institute of Medical Sciences, said that "approaches like this that use computer vision and/or machine learning to detect subtle vascular features predictive of future heart health appear promising," but also noted that the abstract's level of detail did not allow him to draw firm conclusions on whether the work is superior to other recent attempts to predict heart disease risk.
Ware also noted that the genetic risk scores that the Edinburgh team used are not typically available for patients – although they could be."It will be interesting to see whether the model identifies people at risk of [myocardial infarction] without knowing their polygenic risk score, as this would be simpler to implement in practice," he said. But "genetic risk scores also promise to be very powerful tools for early identification of at-risk individuals, and indeed genetic risk can be assessed from birth."