Predicting Risk of Cardiovascular Disease Using Retinal Optical Coherence Tomography Imaging

Core Concepts

Retinal optical coherence tomography (OCT) imaging, combined with self-supervised deep learning and multimodal classification, can effectively predict the risk of future cardiovascular disease events, outperforming the clinically established QRISK3 score.

Abstract

The study investigates the potential of retinal optical coherence tomography (OCT) imaging as an additional tool to predict future cardiovascular disease (CVD) events, such as myocardial infarction and stroke.
The key highlights are:

The researchers employed a self-supervised deep learning approach using Variational Autoencoders (VAE) to learn low-dimensional latent representations from 3D OCT images, capturing characteristics of different retinal layers.

These latent features, combined with patient demographic and clinical data, were used to train a Random Forest (RF) classifier to differentiate between patients at risk of future CVD events and the control group.

The self-supervised VAE feature selection and multimodal RF classifier achieved an AUC of 0.75, outperforming the clinically established QRISK3 score (AUC = 0.597).

The choroidal layer visible in OCT images was identified as an important predictor of future CVD events using a novel approach to model explainability.

Retinal OCT imaging provides a cost-effective and non-invasive alternative to predict the risk of cardiovascular disease, with potential for widespread application in optometry practices and hospitals.

Stats

Retinal OCT imaging data from 630 patients who suffered acute myocardial infarction or stroke within 5 years, and an equal number of control participants without CVD.
Demographic and clinical data, including age, gender, ethnicity, BMI, blood pressure, and HbA1c levels.

Quotes

"Retinal OCT imaging provides a cost-effective and non-invasive alternative to predict the risk of cardiovascular disease and is readily accessible in optometry practices and hospitals."
"The choroidal layer visible in OCT images was identified as an important predictor of future CVD events using a novel approach to model explainability."

Key Insights Distilled From

Predicting risk of cardiovascular disease using retinal OCT imaging

by Cynthia Mald... at arxiv.org 03-29-2024

https://arxiv.org/pdf/2403.18873.pdf

Deeper Inquiries

How can the proposed approach be further improved by incorporating additional retinal imaging modalities, such as fundus photography and OCT angiography?

Incorporating additional retinal imaging modalities, such as fundus photography and OCT angiography, can enhance the predictive performance of the proposed approach in several ways. Fundus photography provides a two-dimensional depiction of the retina, offering complementary information to the 3D imaging capabilities of OCT. By combining data from fundus photography with OCT images, the model can capture a more comprehensive view of the retinal structure and vasculature, potentially improving the accuracy of risk prediction.
OCT angiography, on the other hand, allows for visualization of the retinal and choroidal vasculature without the need for contrast agents. By integrating information from OCT angiography, the model can assess the microvascular perfusion and flow patterns in the retina, providing additional insights into vascular health and potential cardiovascular risk factors.
Furthermore, by leveraging a multi-modal approach that includes data from different imaging modalities, the model can capture a broader range of features and patterns that may be indicative of cardiovascular disease risk. This comprehensive analysis can lead to a more robust and accurate predictive model.

What are the potential limitations of using spectral-domain OCT, and how might the use of swept-source OCT with improved choroidal visualization impact the predictive performance?

Spectral-domain OCT (SD-OCT) has limitations in terms of its ability to visualize deeper layers of the retina, particularly the choroid, due to light scattering issues. This can impact the resolution and clarity of images obtained using SD-OCT, potentially limiting the model's ability to capture detailed information about the choroidal layer, which has been identified as an important predictor of cardiovascular disease risk.
In contrast, swept-source OCT (SS-OCT) offers improved depth penetration and enhanced visualization of deeper tissue layers, including the choroid. By utilizing SS-OCT with improved choroidal visualization, the model can access more detailed and accurate information about the choroidal morphology and microvasculature. This enhanced imaging capability can provide a more comprehensive assessment of the retinal features associated with cardiovascular disease risk, potentially leading to better predictive performance.
The use of SS-OCT can overcome the limitations of SD-OCT in visualizing the choroid and other deep retinal layers, allowing for a more detailed analysis of the structural and vascular changes that may be indicative of cardiovascular risk. This improved imaging technology can enhance the model's ability to identify subtle abnormalities and patterns in the retina that are relevant for predicting cardiovascular disease.

Could the insights gained from the retinal features associated with cardiovascular disease risk provide new avenues for understanding the underlying pathophysiological mechanisms linking the eye and the cardiovascular system?

The insights gained from analyzing retinal features associated with cardiovascular disease risk can indeed provide new avenues for understanding the underlying pathophysiological mechanisms linking the eye and the cardiovascular system. The retina is considered a window to systemic health, and changes in retinal morphology and vasculature have been linked to various cardiovascular conditions.
By identifying specific retinal features that are predictive of cardiovascular disease risk, such as choroidal thinning or abnormalities in the inner retinal layers, researchers can uncover potential biomarkers that reflect systemic vascular health. These retinal biomarkers may indicate early signs of microvascular dysfunction or vascular pathology that precede clinical manifestations of cardiovascular disease.
Understanding the relationship between retinal features and cardiovascular risk can shed light on the shared pathophysiological pathways between the eye and the cardiovascular system. For example, changes in retinal microvasculature may reflect systemic endothelial dysfunction or vascular remodeling, providing insights into the early stages of cardiovascular disease development. By elucidating these mechanisms, researchers can potentially identify novel targets for early intervention and prevention strategies for cardiovascular disease.