AI-Enabled Imaging of Retinal Vascular Networks Can Predict Cardiovascular Disease/Death

AI-powered imaging of the retinal vein and artery network can accurately predict cardiovascular disease and death without the need for blood tests or blood pressure measurements, study reveals¹ Publish online at British Journal of Ophthalmology.

The findings pave the way for a highly effective, non-invasive screening test that does not require a clinic for people at moderate to high risk of cardiovascular disease, the study said in a news release. person suggests.

Cardiovascular disease, including cardiovascular disease, coronary heart disease, heart failure and stroke, is the leading cause of ill health and death worldwide, accounting for one in four deaths in the UK alone. increase.

Several risk frameworks exist, researchers say, but they don’t always pinpoint exactly who will develop cardiovascular disease or die.

Previously published research suggests that the width of the small veins and arteries (arterioles and venules) of the retina (the parts of the eye that receive and organize visual information) may provide an accurate early indicator of cardiovascular disease. indicates that there is

However, it is not clear whether these findings apply consistently and equally to men and women.

Researchers therefore developed a fully automated artificial intelligence (AI)-enabled algorithm (Quantitative Analysis of Retinal Vessel Topology and Size, or QUARTZ for short) to predict retinal vasculature imaging feasibility and vascular health. developed a model to assess known risk factors for death.

According to a release, the researchers applied QUARTZ to retinal images from 88,052 UK Biobank participants aged 40 to 69 years, specifically measuring retinal arteriole and venule width, vessel area, and curvature (twist). We examined the extent and developed a predictive model. Death from stroke, heart attack, and cardiovascular disease.

In a news release, they noted that these models were then applied to retinal images of 7411 participants aged 48 to 92.

The performance of QUARTZ was compared individually and jointly with the widely used Framingham Risk Scores framework.

All participants’ health status was followed for an average of 7–9 years, during which time 327 died of cardiovascular disease among 64,144 UK Biobank participants (mean age 56 years) and 5862 EPIC-Norfolk participants 201 died of cardiovascular disease (mean age 67 years). ).

In men, arteriole and venule width, tortuosity, and width variability emerged as significant predictors of cardiovascular disease mortality. In females, variability in arteriole and venule area and width and venule tortuosity and width contributed to risk prediction.

The predictive impact of retinal vasculature on cardiovascular disease mortality interacted with smoking, antihypertensive drugs, and previous heart attack.

Overall, these predictive models based on age, smoking, medical history, and retinal vasculature capture half to two-thirds of the most at-risk cardiovascular disease deaths.

The retinal vasculature model also has approximately 5% more stroke cases in UK Biobank men, 8% more cases in UK Biobank women, and 3% more at-risk EPIC-Norfolk men. We caught a lot of cases, but about 2% fewer cases in EPIC-Norfolk women. And the Framingham Risk Scores captured more heart attack cases among those most at risk.

Adding retinal vasculature to the Framingham Risk Scores had only a modest effect on stroke or heart attack prediction. However, a simpler, noninvasive risk score based on age, sex, smoking, medical history, and retinal vasculature, without the need for blood tests or blood pressure measurements, was performed similarly to the Framingham risk score, the researchers said. says.

They acknowledge that both groups of study participants had healthier lifestyles than other geographically similar middle-aged people. However, this is the largest population-based study of retinal vasculature, and predictive models have been externally validated in a large number of people, they explain.

Retinal imaging is already a popular method in the UK and US, researchers note.

“AI-enabled vascular instrumentation risk prediction is fully automated, low-cost, non-invasive, and does not depend on ‘high street’ availability and blood sampling or [blood pressure measurement] is not required,” the researchers concluded.

Furthermore, researchers [retinal vasculature] is a microvascular marker, [heart attack] Except perhaps in women, this is the larger vessel.

“In the general population, it can be used as a non-contact form of systemic vascular health check to triage individuals at moderate to high risk of cardiovascular mortality for further clinical risk assessment and appropriate intervention.” they said.

For example, in the UK, they suggest it could be included in NHS health checks for primary care between the ages of 41 and 74.

In a linked editorial, Ify Mordi (MPH, PMP) and Emanuele Trucco (PhD), University of Dundee, Scotland, say that using changes in retinal vasculature to inform overall cardiovascular risk is “certainly attractive.” described as “intuitive and intuitive” yet still become part of routine clinical practice.

“Using retinal screening in this fashion will probably require a significant increase in the number of ophthalmologists or other trained assessors,” they suggest, suggesting which specialties are available for preventive treatment and additional intervention. They are asking who will be responsible for the workload of: ophthalmologists, cardiologists, or primary care physicians?

“What is needed now is ophthalmologists, cardiologists, primary care physicians and computer scientists working together to design studies and use this information to determine whether clinical outcomes improve. If so, we will work with regulatory bodies, scientific bodies and health systems to optimize clinical workflows and enable practical implementation in routine practice,” they conclude.

reference

1 Alicia R. Ridnicka, Roshan A. Welikala, Sara A. Barman, Paul J. Foster, Robert Luben, Shabina A. Hayat, Kay-Tee Khaw, Peter H. Whincup, David p. Strachan, Christopher G. Own, UK Biobank Eye and Vision Consortium. Artificial intelligence-enabled retinal vascular measurements for prediction of cardiovascular mortality, myocardial infarction, and stroke. British Journal of OphthalmologyPublished October 2022 DOI: https://doi.org/10.1101/2022.05.16.22275133