Application value of artificial intelligence-basedretinal microvascular analysis in diagnosis of diabetes complications.

Objective To explore the application value of artificial intelligence (AI) based retinal microvascular analysis in the diagnosis of diabetes retinopathy (DR), diabetes nephropathy (DN) and other diabetes complications. Methods From January to December 2022, 305 subjects from the Health Medicine Department and Endocrine Department of the Eastern Theater Command General Hospital of PLA were divided into healthy control group (n = 119), diabetes without DR group (n = 100), and diabetes with DR group (n = 86) according to the condition of diabetes and its complications. The group of diabetes with DR was further divided into the group without DN (n = 50) and the group with DN (n = 36). Clinical data, laboratory test indicators, and fundus photography results of all the subjects were collected. Independent sample t-test, Kruskal Wallis H-test, and logistic regression analysis were used for data analysis. Results There were statistically significant differences in FBG, 2hPBG, HbA1c, central retinal artery equivalent (CRAE), and retinal arteriovenous ratio (AVR) among the three groups (all P < 0.05). In patients with diabetes complicated with DR, there were statistically significant differences in the history of diabetes, total number of bleeding, total area of exudation, maximum area of exudation and total number of exudation between non-DN group and DN group (all P < 0.05). There was correlation between retinal microvascular indicators and diabetic retinopathy. When the total area of retinal hemorrhage, total area of exudation, and maximum area of exudation increased, the risk of diabetic retinopathy increased. Conclusion AI-based retinal microvascular analysis has value in providing prompt indication and assisting with the diagnosis of complications in diabetic patients. [ABSTRACT FROM AUTHOR]