Approved Research
Deep Learning based systemic disease screening at the population level using retinal images
Lay summary
The eye is an important organ of the body that manifests changes in response to several systemic diseases. Several studies have attempted to characterize the retinal changes that occur in response to specific systemic disorders so that these changes may be used as markers for systemic diseases. Visualization of retinal alterations can be done by obtaining retinal photographs of patients that preclude the need for complex and/or invasive diagnostic procedures.
Systemic diseases are those that affect the entire body rather than just one organ or tissue. Such disorders include diabetes mellitus, hypertension, chronic kidney disease, rheumatoid arthritis, atherosclerosis, and metabolic syndrome. Most systemic diseases such as diabetes, arthritis, hypertension, and kidney disease are chronic and result in severe health consequences for the patients. Early detection of such conditions is important for optimal management through dietary and lifestyle changes resulting in better health outcomes for the patient.
The aim of this project is to characterize the retinal alterations associated with systemic diseases so that they may be used as early predictors of the disease. As the burden of systemic diseases is increasing globally, we need effective, accurate and cost-effective approaches to determine the susceptibility of patients to certain conditions. We also intend to characterize the underlying genetic changes that are associated with retinal changes and systemic diseases so that we can investigate important associations between retinal and genetic changes of a specific systemic condition.
Furthermore, this project will aim to develop a novel deep neural network (DNN) model to predict the risk of systemic diseases using both retinal images and genomic data. This model will consist of two parts: a classification network to identify high-risk patients and a regression network to assign a risk score. This will allow for more accurate and individualized predictions, taking into account both retinal and genetic factors.
Additionally, the project will include a population-based approach, including individuals from different ethnic backgrounds, to account for variations in retinal pigmentation and genetic makeup. This will ensure that the developed models are generalizable and applicable to a diverse population.
The ultimate goal of this project is to develop a non-invasive and cost-effective diagnostic tool for early detection of systemic diseases using retinal images and genomic data. The results of this study have the potential to greatly improve the management and outcome of chronic conditions and improve the quality of life for patients.