Principal Investigator: Dr Wei Wang
Department: Sun Yat-Sen UniversityTags: 62491, eye, kidney, Machine Learning
Aims: The overall goal of this project is to develop and validate a screening test for predicting chronic kidney diseases (CKD) through ocular imaging. Specific aims are: 1) To explore the associations between major eye diseases and a wide range of CKD. 2) To explore the associations between retinal biomarkers derived from retinal imaging, and CKD, as well as specific measures of kidney functions. 3) To develop and validate the clinical application of deep learning to ocular imaging for CKD, and kidney function prediction.
Scientific rationale: Aging populations have inevitably led to a substantial rise in prevalence of age-related disease, including CKD. Accurate prediction of an individual risk of a disease is the foundation for targeting preventive treatment at persons who are asymptomatic but at high risk of the disease. However, currently available clinical prediction tools are limited by the lack precision, involvement of invasive procedures, high costs and have other limitations.
The eye shares striking structural, developmental, physiological, and pathogenic pathways with kidney. Over the past decades, the research on this field has been developing rapidly and novel ocular biomarkers have been identified to be associated with CKD. Nevertheless, much of the previous work has focused on a few features explicitly detected by ophthalmic expert using the standardized reviewing systems. Therefore, “personalized” image features and/or patterns have not yet fully investigated.
As a special type of deep-learning technique that has been optimized for images, deep learning approaches have achieved state-of-the-art performance in prediction models across several domains. The ability of learning the “personalized” image features and/or patterns for the deep learning technology may improve the prediction of CKD based on retinal imaging.
Duration of the project: 36 months.
Public health impact: The heavy public burden of the leading causes of morbidity and mortality due to CKD calls for robust and feasible prediction models to identify high-risk individuals. Our proposed project, to develop and validate a efficient screening modality for predicting CKD incorporating deep learning techniques based on ocular imaging, aims to obtain personalized systemic diseases’ risk information. Thus our program has great potential to improve prediction precision. Furthermore, the low cost, easy accessibility and transfer, as well as non-invasive features make ocular imaging suitable for large scale population screening. Thus, our program will help facilitate preventive strategies targeting at the right people and improve clinical outcomes, in turn reducing the burden of CKD.