Skip To Content


SNUH’s new AI model assesses cardiovascular death risk

Hit : 214 Date : 2020-04-24

A research team at Seoul National University Hospital (SNUH) said Friday that it has developed an artificial intelligence (AI) model that can find atherosclerosis.

The research team, led by Professor Park Sang-min, used 15,408 fundus photographs that have been accumulated for 12 years at the Center for Health Promotion and Optimal Aging to develop the model.

The images of eyes, which the AI model used to predict arteriosclerosis by analyzing the blood vessels and anatomical structures of the eye (Credit: Seoul National University Hospital)

Atherosclerosis is a disease, which narrows the inside of an artery as a result of the buildup of plaque and cell proliferation, leading to atheroma, it said. The disease causes complications such as stroke, myocardial infarction, and peripheral vascular disease.

Early detection of atherosclerosis is vital for the patients because they generally do not show symptoms, but the diagnostic imaging requires high cost.

“Eyes are the only organs, in which medical professionals can directly see the people’s vascular health,” Professor Park said. “Previously, we could see carotid hardening only through carotid ultrasonography. Now, however, we can accurately evaluate carotid stenosis by using AI-based diagnostic methods on the fundus image.”

The research team analyzed the effect of AI fundus arteriosclerosis scores on the risk of death by cardiovascular problems in a group of 30,227 patients.

The results showed that even when the cardiovascular Framingham Risk Score, which evaluates the risk of cardiovascular disease, was calibrated and compared, the score independently has a close relationship to the risk of cardiovascular death. If the AI ​​fundus arteriosclerosis score belonged to a high-risk group, the rate of cardiovascular death increased by eightfold.

The study also has significant clinical value because the testers can distinguish and predict the risk of cardiovascular death in patients with moderate cardiovascular risk, who require additional tests to decide whether to treat them more aggressively or not.

“AI fundus arteriosclerosis model predicts arteriosclerosis by using anatomical factors such as retinal vessels,” said Chang Joo-young, the lead author. “The existing AI study lacked clinical value as it analyzed the fundus image only by the moment when it was taken, but this one is significant in that it predicted future cardiovascular death by forming a retrospective cohort.”

Professor Park revealed plans to develop a new AI-based fundus imaging device, which can quickly and accurately diagnose arteriosclerosis based on their recent attainment.

The article was published in the latest issue of the American Journal of Ophthalmology, published by Elsevier.

<© Korea Biomedical Review, All rights reserved.>


전체 메뉴

전체 검색

전체 검색