Related Experiment Video
Updated: May 23, 2025
Fundus Photography as a Convenient Tool to Study Microvascular Responses to Cardiovascular Disease Risk Factors in Epidemiological Studies
Published on: October 22, 2014
Addressing underestimation and explanation of retinal fundus photo-based cardiovascular disease risk score: Algorithm
Zhihong Lin1, Chi Liu2, Danli Shi3
1The AIM for Health Lab, Monash University, Melbourne, Australia; Faculty of Engineering, Monash University, Melbourne, Australia.
Insights
This study developed a deep learning model to predict cardiovascular disease (CVD) risk from retinal images, improving accuracy and identifying key image features. The model learned risk prediction from age and sex representations, addressing underestimation issues.
Area of Science:
- Ophthalmology
- Cardiology
- Artificial Intelligence
- Medical Imaging
Background:
- Cardiovascular disease (CVD) risk prediction is crucial for preventative care.
- Retinal fundus photographs offer a non-invasive window into systemic vascular health.
- Existing AI models for CVD risk prediction from retinal images may suffer from underestimation.
Purpose of the Study:
- To address the underestimation problem in AI-based CVD risk prediction using retinal images.
- To investigate the underlying mechanisms of a deep learning (DL) model in assessing CVD risk.
- To evaluate the model's performance on diverse datasets and camera types.
Main Methods:
- An ordinal regression deep learning (DL) model was developed for 10-year CVD risk score prediction.
- Transfer learning and saliency maps were employed to explore the DL model's decision-making process.
- Model training utilized UK Biobank data (34,652 participants); validation used Australian datasets (401 participants) with desktop and portable retinal cameras.
Main Results:
- The DL model achieved a mean risk-level accuracy of 0.772 on UK Biobank data.
- Area Under the Receiver Operating Characteristic (AUROC) curves for moderate and high risk were 0.849 and 0.874, respectively.
- External validation showed risk-level accuracies of 0.715 (desktop camera) and 0.656 (portable camera).
Conclusions:
- The proposed DL model successfully minimized the underestimation of CVD risk.
- The model primarily learned CVD risk from age- and sex-related image features.
- Peripheral small vessels and macula areas were identified as image features associated with high CVD risk.
Objective:
To resolve the underestimation problem and investigate the mechanism of the AI model which employed to predict cardiovascular disease (CVD) risk scores from retinal fundus photos.
Methods:
An ordinal regression Deep Learning (DL) model was proposed to predict 10-year CVD risk scores. The mechanism of the DL model in understanding CVD risk was explored using methods such as transfer learning and saliency maps.
Results:
Model development was performed using data from 34,652 participants with good-quality fundus photographs from the UK Biobank and a dataset for external validation collected in Australia comprised of 1376 fundus photos of 401 participants with a desktop retinal camera and a portable retinal camera. The mean [SD] risk-level accuracies across cross-validation folds was 0.772 [0.008], while AUROC for over moderate risk was 0.849 [0.005] and the AUROC for high risk was 0.874 [0.007] on the UK Biobank dataset. The risk-level accuracy for images acquired with the desktop camera data was 0.715, and the accuracy for portable camera data was 0.656 on the external dataset.
Conclusions:
The DL model described in this study has minimized the underestimation problem. Our analysis confirms that the DL model learned CVD risk score prediction primarily from age- and sex-related image representation. Model performance was only slightly degraded when features such as the retinal vessels and colours were removed from the images. Our analysis identified some image features associated with high CVD risk status, such as the peripheral small vessels and the macula areas.
