We reported the development and validation of deep-learning models for the automated measurement of retinal vessel calibre in retinal photographs, by using diverse multi-ethnic multi-country datasets amounting to more than 70,000 images, including data from UK Biobank. Retinal vessel calibres measured by the models and by expert human graders showed high agreement, with overall intraclass correlation coefficients between 0.82 and 0.95. The models performed comparably to or better than expert graders in associations of measurements of retinal vessel calibre with cardiovascular disease (CVD) risk factors, including blood pressure, body mass index, total cholesterol and glycated-haemoglobin levels. In retrospectively measured prospective datasets from a population-based study, baseline measurements by the deep-learning system were associated with incident CVD. Our findings motivate the development of clinically applicable explainable end-to-end deep-learning systems for the prediction of CVD on the basis of features of retinal vessels in retinal photographs.
Risk Factors, Biomarkers and Impact of Major Vision-Threatening Eye Diseases
Aims: The primary objective of this research is to evaluate biologic, lifestyle factors and biomarkers related to the development of major vision-threatening eye diseases in Asian and western populations, using data from the SEED study and UK Biobank study. We will further utilize the collected clinical data and images to develop a novel artificial intelligence-based system to predict the onset and progression of diabetic retinopathy or other major eye diseases, using the UK Biobank data for validation.
Scientific rationale: The prevalence of vision-threatening eye diseases is projected to increase steadily with an ageing population. The loss of vision is often irreversible, resulting in loss of productivity, inability to perform activities of daily living, and a substantial reduction in quality-of-life. This poses a significant challenge to healthcare professionals of today and tomorrow. Thus, there is an urgent need for healthcare systems in the world to transit from a curative paradigm to one that emphasizes on prevention. This is especially so in eye care as the gap between supply and demand of eye care services widen globally.
This transition in healthcare system needs to be informed by large-scale genomic and detailed clinical data from well-designed population-based prospective cohort studies, and the UK Biobank is an excellent example of such a resource-rich avenue. The UK Biobank will provide us with the data breadth and depth needed to gain a deeper insight into the individual and combined effects of genetic and environmental determinants for a wide range of eye diseases in adult population. These diseases may be caused by many different exposures, each with a modest effect and interact with one another in complex ways. Therefore, a large sample size is needed to study the specific effects of any specific exposure. The extensive dataset from the UK Biobank will allow in-depth analysis and comparison to be made between different populations.
Project duration: 36 months
Public health impact: Findings from this project may provide important information on risk factors of vision-threatening eye diseases, which may improve our understanding of the complex gene-environment mechanism involved in disease development or progression, and accelerate the development of prevention or early detection program for major eye diseases. This project will also provide useful information on the generalizability of risk factors and impact of major eye diseases in genetically different populations.
|Lead investigator:||Professor Ching-yu Cheng|
|Lead institution:||Singapore Eye Research Institute|
|3154||Prediction of systemic biomarkers from retinal photographs: development and validation of deep-learning algorithms||Rim TH et al.||2020||Lancet Digit Health|