Applying a convolutional neural network as a statistical method to evaluate the correlation between diabetic retinopathy and diabetic nephropathy

Abstract

Abstract Objective At present, there is no consensus on the relationship between diabetic nephropathy (DN) and diabetic retinopathy (DR), and there is a lack of imaging data to prove the correlation between them. Therefore, we aimed to investigate the common risk factors for DN and DR in patients with type 2 diabetes and used Emission Computed Tomography(ECT) imaging technology and clinical test criteria to assess the correlation between DR and DN.Purpose The convolutional neural network was employed to observe and detect pathological changes in DR and DN images. And further analyze the correlation between DR and DN through clinical test data.Results In this study, we discovered that with the aggravation of DR, SCR, BUN and ACR gradually increased while GFR decreased. The incidence of DN in the non-DR, mild-NPDR, moderate-NPDR, severe-NPDR and PDR groups was 4.17%, 28.33%, 55%, 75.83% and 91.67%, respectively. Multivariate linear regression analysis showed that duration of T2D, smoking, HbA1c, TC, TG, HDL-c, LDL-c, UAlb, Scr, BUN, UAlb, ACR, GFR, ACR and GFR were independent risk factors for DR. Renal dynamic ECT imaging analysis demonstrated that with the aggravation of DR, renal blood flow perfusion gradually decreased, thus resulting in a decrease in renal filtration function. In T2D patients, DR and DN show a linear aggravation relationship, and hypercholesterolemia and renal dysfunction are common risk factors for DR and DN.Conclusion The convolutional neural network provides a more accurate, efficient and easier way to analyze DR and DN images. Early screening of the renal function index in DR patients using ECT imaging technology will help to identify and prevent DN as early as possible.