Approved Research

An artificial intelligence-based approach for the early detection of chronic kidney diseases from digital retinal photographs

Lay summary

Chronic Kidney Disease (CKD) is a serious condition that can lead to kidney failure. Diabetes is the leading cause of CKD, and people with diabetes are advised to get regular eye exams to monitor for signs of retinopathy, which is linked to CKD. To make an early diagnosis of CKD easier, we can use artificial intelligence to analyze retinal images and predict the onset and stage of CKD. We will use deep-learning models to analyze the retinal images and collect clinical data from patients, such as age, blood pressure, and diabetes status, to predict their kidney function. We will use this information to train separate models to predict kidney function and blood glucose levels. Then, we will use these predicted values to detect CKD and its stages. Finally, we will combine the retinal image data and clinical information to create a more accurate model. Early detection is appealing because inexpensive intervention can ameliorate renal function early and prevent continuous deterioration. The shortcomings of traditional analysis methods make a new comprehensive assessment approach urgently needed in kidney disease patients and a large population scale. ML and DL combined with retinal imaging is a relatively prominent frontier in this research area.

Scope extension, June 2024:

Chronic kidney disease (CKD) is a highly consequential medical condition that poses significant morbidity and healthcare expenditures, given that patients typically remain asymptomatic for extended periods before clinical manifestations appear. By the time the condition is detected, it may have progressed to a degree that renders successful intervention difficult. At an advanced stage, CKD can only be managed through dialysis or transplantation, which can be prohibitively expensive. However, key risk factors can generally be managed to prevent CKD from developing. Accordingly, the present investigation aims to develop an artificial intelligence (AI)-based model utilizing retinal tomography to detect and monitor the progression of chronic kidney disease.

In the extended scope, we aim to look into kidney function and the factors or organs that are associated with it. We strive to utilize the different features available in UK Biobank that include but are not limited to electrocardiogram (ECG), magnetic resonance images (MRI), early life factors, lifestyle, and genetic data (e.g., polygenic risk score, Telomeres). Using various features, we will aim to pinpoint risk factors associated with kidney function. Also, we will investigate how kidney function can affect other organs, such as the heart, liver, and brain. To achieve this aim, we plan to use the necessary software (e.g., MATLAB, python) to perform different statistical analyses (e.g., Cox proportional hazard model).