Weaning from mechanical ventilation: Feature extraction from a statistical signal processing viewpoint

Clinicians' decision for mechanical aid discontinuation is a challenging task that involves a complete knowledge of a great number of clinical parameters, as well as its evolution in time. Respiratory pattern variability appears as a useful extubation readiness indicator, and thus can be used as an informative feature in a statistical pattern recognition framework. Reliable assessment of this variability involves a set of signal processing techniques that should be carefully evaluated for statistical validity. This paper evaluates different variability extraction techniques aimed to build a Bayesian classifier for weaning readiness decision. As a conclusion, Sample Entropy is selected as the best performance extraction method. By calculating it over tidal volume signals, and with mean respiratory rates as additional input patterns, a 2D Bayesian classifier is constructed with principal component analysis selection. The obtained misclassification probability (P
_e
= 0.2141) is acceptable if compared with performance of single feature classifiers.