Predicting cardiovascular disease risk using retinal optical coherence tomography imaging.

Area of Science:

• Ophthalmology
• Cardiology
• Artificial Intelligence

Background:

• Cardiovascular Diseases (CVD) are a leading global cause of mortality.
• Non-invasive imaging is vital for early CVD detection and prevention.
• Optical Coherence Tomography (OCT) detects microvascular changes, aiding early identification of at-risk patients.

Purpose of the Study:

• To investigate the potential of retinal OCT as an additional imaging technique for predicting future CVD events.
• To utilize deep learning for extracting features from OCT images to assess CVD risk.

Main Methods:

• Analysis of retinal OCT data from 2,846 UK Biobank participants (612 with future CVD events, 2,234 controls).
• Application of a self-supervised deep learning (Variational Autoencoder) approach to derive latent representations from 3D OCT images.
• Training a Random Forest classifier using latent features and clinical data to predict myocardial infarction (MI) or stroke.

Main Results:

• The predictive model achieved an Area Under the Curve (AUC) of 0.75, with 0.70 sensitivity, 0.70 specificity, and 0.70 accuracy.
• The choroidal layer in OCT images was identified as a significant predictor of future CVD events via model explainability.

Conclusions:

• Retinal OCT imaging, augmented by deep learning, shows promise as a predictive tool for cardiovascular events.
• This approach can help identify individuals at increased risk for CVD, facilitating timely intervention.