AI-enabled retina imaging could predict cardiovascular disease, death

Computer Heart Social

U.K. researchers have reported on the use of artificial intelligence (AI)-enabled imaging of the retina to predict cardiovascular disease and death without the need for blood tests or blood pressure measurement.

The researchers, describing the system recently in the British Journal of Ophthalmology, said it uses images of the retina’s network of veins and arteries to evaluate circulatory mortality, myocardial infarction, and stroke.  

Although further evidence of its applicability and benefit is needed, it could enable the development of a noninvasive test to screen people who are at medium to high risk of circulatory disease, the researchers said.

“A natural progression of our work would be to conduct a clinical trial to evaluate the utility in terms of changes in health outcomes such as heart attacks, strokes, and mortality before such a system would be considered for implementation into any healthcare setting,” Alicja Rudnicka, the study’s lead author and professor at the University of London, said in an email interview.

Rudnicka added that the method addresses an unmet need in the U.K., where there is low attendance for cardiovascular risk screenings.

“Having a low cost, accessible, noninvasive screening test in the community to encourage clinical risk assessment … is highly likely to help prolong disease-free status in an ever-aging population with increasing comorbidities and assist with minimizing healthcare costs associated with lifelong vascular diseases,” she added.

Circulatory diseases, including cardiovascular disease, coronary heart disease, heart failure, and stroke, are major causes of ill health and death worldwide. While several risk frameworks exist, physicians still have trouble using them to identify who will go on to develop or die from these diseases.

Previous research has demonstrated that “narrow retinal arterioles show a clear association with higher blood pressure (BP), hypertension, and with incident CVD,” the researchers wrote.

“Prospective associations have been largely based on retinal vessel width with mortality, incident stroke, and with [coronary heart disease] (in women, not men), from restricted measurement areas of the retina,” the study authors said, adding, “Measurements are often not automated, requiring operator involvement, which limits application to large populations.”

The U.K. team’s automated AI-enabled algorithm is called QUantitative Analysis of Retinal vessels of Topology and siZe, or QUARTZ. In the study, they used QUARTZ to develop models to assess the potential of retinal vasculature imaging to predict vascular health and death.

They applied the algorithm to 88,052 UK Biobank participants between the ages of 40 and 69, looking specifically at the width, vessel area, and degree of curviness (tortuosity) of the arterioles and venules in the retina to develop prediction models for stroke, heart attack, and death from cardiovascular disease.

The team subsequently applied these models to the retinal images of 7,411 participants, ages 48 to 92, in the European Prospective Investigation into Cancer (EPIC)-Norfolk study.

The performance of QUARTZ was also compared with the widely used Framingham Risk Scores framework.

The health of all participants was tracked for an average of seven to nine years, during which time there were 327 circulatory disease deaths among 64,144 UK Biobank participants (average age 56) and 201 circulatory deaths among 5,862 EPIC-Norfolk participants (average age 67).

In men, arteriolar and venular width, tortuosity, and width variation emerged as important predictors of death from circulatory disease. In women, arteriolar and venular area and width and venular tortuosity and width variation contributed to risk prediction. 

Although the system shows promise, the team acknowledged the study's limitations. Their chosen population led healthier lifestyles than other geographically similar age groups, and the population was largely white.

“Confirmation of model performance in other cohorts with higher CVD rates and in different (especially non-white) ethnic groups would be informative,” they wrote. 

Page 1 of 2
Next Page