The artificial intelligence supported system in the diagnosis and risk prediction of metabolic disorders based on multimodal data

Metabolic diseases, including diabetes, hypertension, dyslipidemia, have become key to global public health and chronic disease management, bringing huge social burdens. Early diagnosis and intervention of metabolic disorders and their complications are key to reducing disability and death. However, metabolic disorders often have no significant clinical symptoms and manifestations in the early stages, or only some nonspecific symptoms. Therefore, how to identify patients with metabolic abnormalities in the community population, identify high-risk groups and then carry out targeted interventions for long-term complications and mortality risks is an urgent problem to be solved. The retina is the only part of the body where blood vessels and nerve tissue can be directly observed by noninvasive examination, and there are a large number of imaging features for understanding the comprehensive situation and systemic diseases of patients. The use of deep learning technology can accurately detect the patient’s age, sex, smoking, blood pressure and other cardiovascular risk factor information from the fundus image alone, and a number of subsequent studies have further confirmed that the fundus picture contains a large amount of biological information that is difficult to be recognized by the naked eye or traditional image recognition technology.
We have the aim to study the use of deep learning technology combined with fundus images, clinical indicators and other multimodal information to identify and assess metabolic diseases such as diabetes, hypertension, dyslipidemia, chronic kidney disease and other metabolic diseases and their complications and target organ damage, and further predict the disease progression of diabetic microvascular complications (diabetic retinopathy, diabetic nephropathy.), hypertension and its target organ damage (hypertensive retinopathy, arteriosclerosis, hypertensive nephropathy.), myopic maculopathy, and dyslipidemia-related diseases (coronary heart disease, cerebrovascular disease, metabolism-related fatty liver disease.) based on the UK Biobank data.
We therefore anticipate that our analyses with the UK Biobank data will contribute to the early diagnosis and intervention of metabolic disorders and their complications ,thus reducing disability and death. This would be of high value to global public health and chronic disease management.