Using deep learning to analyze retinal photographs for cardiovascular disease prediction

Prediction of Incident Atherosclerotic Cardiovascular Disease From Retinal Photographs Via Deep Learning

This study is testing whether a new computer program can use pictures of the back of the eye to predict the risk of heart disease in people over five years.

Quick facts

• Study type: Observational
• Enrollment: 2400 (estimated)
• Ages: 20 Years to 79 Years
• Sex: All
• Sponsor: Yonsei University (Academic / other)
• Locations: 1 site (Seoul)
• Trial ID: NCT04749927 on ClinicalTrials.gov

What this trial studies

This observational study aims to validate a deep learning algorithm that predicts cardiovascular disease risk based on fundus photographs. Participants will have their retinal images taken to assess various cardiovascular conditions, including myocardial infarction and heart failure. The study will follow participants for five years to track major cardiovascular events and verify the algorithm's predictive capabilities. Fundus photographs will be collected twice in the first year and again two years later to enhance the data set for analysis.

Who should consider this trial

Good fit: Ideal candidates include individuals diagnosed with myocardial infarction, heart failure, or high-risk subclinical atherosclerosis.

Not a fit: Patients without any cardiovascular conditions or those who do not meet the specific inclusion criteria may not benefit from this study.

Why it matters

Potential benefit: If successful, this study could lead to improved early detection and risk assessment of cardiovascular diseases using non-invasive retinal imaging.

How similar studies have performed: While the use of deep learning in medical imaging is gaining traction, this specific approach to predicting cardiovascular disease risk from retinal photographs is relatively novel.

Eligibility criteria

Inclusion Criteria:

1. Myocardial infarction (Patients diagnosed with myocardial infarction within 5 years and confirmed significant coronary artery stenosis by cardiovascular angiography)
2. Heart failure with reduced EF (\<40% of LVEF on echocardiography or magnetic resonance imaging)
3. Heart failure with preserved EF (≥40% of LVEF on echocardiography and NT-proBNP ≥200 pg/mL and LAVI ≥ 34 ml/m2 or LVMI ≥115 g/m2 (men) or LVMI ≥95 g/m2 (women))
4. High risk subclinical atherosclerosis (no symptom and ≥50% stenosis of coronary artery on coronary angio CT or asymptomatic PAOD or cerebral aneurysm or ≥50% stenosis of cerebral artery or ABI \<0.9 or ≥2mm of atherosclerotic plaque or hypoechogenic plaque on carotid ultrasound)
5. Hypertension with target organ damage (proteinuria \[urine albumin/creatinine ratio ≥ 30 mg/g or protein/creatinine ratio ≥ 150 mg/g or 24 hour urine albumin ≥30mg/day or 24 hour urine protein ≥ 150mg/day\] or LV hypertrophy \[on EKG or echocardiography\] or cfPWV \> 10 m/sec or baPWV \> 1800 cm/sec or eGFR \< 60 ml/min/1.72 m2 or atherosclerotic cardiovascular disease or white matter hyperintensity on brain MRI)
6. High risk dyslipidemia (LDL-cholesterol \>190 mg/dL or \> 160 mg/dL inspire of use of moderate or high intensity statin)
7. Diabetes (Type 2 diabetes with more than 5 years of diagnosis or type 1 diabetes with more than 10 years of diagnosis)
8. Low risk (Hypertension that does not meet the above criteria and is controlled by 3 drugs or less 2) Dyslipidemia that does not meet the above criteria and is controlled below the target LDL)

Exclusion Criteria:

1. Serious eye diseases that make it impossible to take adequate quality fundus photography
2. If the subject cannot read and sign the consent form in person

Where this trial is running

Seoul

• Yonsei University College of Medicine — Seoul, South Korea (Recruiting)

Study contacts

• Principal investigator: Sungha Park — Severance Hospital
• Study coordinator: Sungha Park

Conditions Cardiovascular Disease