Support Vector Machine Classification applied on Weaning Trials Patients
B.F. Giraldo (Technical University of Catalonia, Spain), A. Garde (CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Spain), C. Arizmendi (Universitat Politècnica de Catalunya, Spain), R. Jané (Institut de Bioenginyeria de Catalunya (IBEC), Spain), I. Diaz (Hospital de la Santa Creu i Sant Pau, Spain) and S. Benito (Hospital de la Santa Creu i Sant Pau, Spain)
Copyright: © 2008
The most common reason for instituting mechanical ventilation is to decrease a patient’s work of breathing. Many attempts have been made to increase the effectiveness on the evaluation of the respiratory pattern by means of respiratory signal analysis. This work suggests a method of studying the lying differences in respiratory pattern variability between patients on weaning trials. The core of the proposed method is the use of support vector machines to classify patients into two groups, taking into account 35 features of each one, previously extracted from the respiratory flow. 146 patients from mechanical ventilation were studied: Group S of 79 patients with Successful trials, and Group F of 67 patients that Failed on the attempt to maintain spontaneous breathing and had to be reconnected. Applying a feature selection procedure based on the use of the support vector machine with leave-one-out cross-validation, it was obtained 86.67% of well classified patients into the Group S and 73.34% into Group F, using only eight of the 35 features. Therefore, support vector machines can be an interesting classification method in the study of the respiratory pattern variability.
Key Terms in this Chapter
Mechanical Ventilation: Breathing via machines.
Spontaneous Breathing: Natural breathing.
Respiratory Pattern Variability: Seeing the breathing pattern.
Support Vector Machine (SVM): A binary classifier based on a supervised statistical learning through examples, and is a new and powerful learning methodology that can deal mainly with nonlinear classification and regression.