Análisis y clasificación del patrón respiratorio de pacientes en proceso de retirada del ventilador mecánico

Abstract

Estimating the optimal time to remove the ventilatory support from a patient in intensive care remains essential in clinical practice. In this work we study the breathing pattern from the respiratory flow signal in the process of weaning considering the following stages: characterization of the signal from the identification of respiratory cycles, respiratory pattern analysis from mathematical modeling of the resulting series, and classification in order to identify patterns of patients with possible success in the process. We analyzed 153 patients classified into three groups: success, failure and reintubated, according to results of T-tube test. The time series for breathing duration, inspiratory time, expiratory time, and shallow breathing index that resulted in significant differences in the mean, model order, first coefficient and final error of prediction were selected. With them we obtained a classification accuracy of 86% (sensitivity 0.84 - specificity 0.86) using a linear classifier discriminate type. Other classifications were analyzed, such as logistic regression and support vector machines.