A Clinical Data-Based Model for Diabetic Retinopathy Screening in Patient Screening

We present an application of machine learning algorithms and Linear Programming Discriminant Analysis (LPDA) in diabetic retinopathy (DR) screening for patient screening based on clinical and demographic data. The algorithms were trained and tested on a dataset of 1460 individuals selected from the National Health and Nutrition Examination Survey (NHANES) database for the years 2005-2020. The methods proved to be effective by several metrics (precision, recall, F1-Score, accuracy, and area under the ROC curve), achieving results above 90%. The variables used in the model were the patient's age at the time of diabetes diagnosis (real and normalized), insulin use (binary), use of oral antidiabetic drugs (binary), and systolic and diastolic blood pressure (real and normalized).