Assessment of cardiovascular risk through a neural network using body composition data. Evaluación del riesgo cardiovascular a través de una red neuronal utilizando datos de la composición corporal.

Abstract

Introduction: The cardiovascular disease (CVD) is the main world cause of morbidity and mortality, continually, are searching early diagnostic strategies with enough sensibility and specificity to be used massively in the population. Objective: To develop an artificial neural network (ANN), to predict the cardiovascular risk using anthropometric variables, age, and habits. Methods: 256 subjects were analyze, aged 16 to 60 years old, the ANN was feed with the variables: age, gender, smoking, fat percentage, visceral fat and muscular percentage, the ANN training was made with 183 subjects (69%), the output variable was the CVD probability to ten years prognosticated using Assessing cardiovascular risk using SIGN guidelines (ASSIGN) scale, the classification was high risk > 10%, moderate risk 5 – 10%, low risk 1 – 4,9%, very low risk < 1%. Results: The model show significate differences in the blood chemical variables and the body composition between groups p < 0.0001. The Area under the curve for the prediction was: High risk = 0.999, Moderate risk = 0.967, low risk = 0.986, very low risk = 0.981. The sensitivity according to risk was between 0.750 y 1.000 and specificity was between 0.833 and 1.000. Conclusion: An ANN that uses as prediction variable the body composition has a high predictive value, besides it is a good tool for the cardiovascular risk estimation and can be used for the screening in different populations.

Keywords: Artificial Intelligence, Risk factors, Body composition, cardiovascular disease