Implementación de algoritmos de inteligencia artificial para la identificación de pacientes diabéticos utilizando los niveles de lípidos en sangre

In recent years the leading cause of death in Mexico is linked to multifactorial diseases, of which diabetes ranks second, only below heart disease, both related to high cholesterol levels and triglycerides in blood. Objective: Classify patients with diabetes using artificial intelligence algorithms previously trained with total cholesterol, HDL, LDL and triglyceride levels. Materials and methods: Descriptors related to blood lipids belong to the Centro Medico Siglo XXI, composed of a sample of 1019. They are considered: Total Cholesterol Levels, HDL, LDH and Triglycerides. The proposed methodology consists of two main stages: training of artificial intelligence algorithms, in which black box models are developed to look for the relationship of the determinants mentioned and the suffering of diabetes in the subjects (presence = 1, absence = 0), and a second stage for the validation of the algorithms, using as a metric the sensitivity and specificity of the algorithms by means of the ROC curve and the area under the curve (AUC). Results: Logistic regression models, decision trees and support vector machine, acquire a value of 0.613 to 0.727 of AUC, being statistically significant for the automatic detection of diabetic patients. Conclusions: The implementation of Artificial Intelligence algorithms, allow the identification of patients with diabetes using blood lipid metrics, for a computer-aided diagnosis.