Graphical Analysis of Respiration in Postoperative Patients Using Self Organising Maps

Abstract

Previous work has shown that the relationships between respiratory pressures and dimensions can be used to investigate how analgesia, airway obstruction and hypoxia are related. These relationships can be more clearly visualised by plotting the different pairs of signals against each other to create a graphical representation of the respiratory mechanisms. This technique of visual classification has been automated using Self Organising Maps. Respiratory signal can be classified into categories on a breath by breath basis without having to explicitly indicate the start of inspiration. Having achieved this we have established a system for extending the research from a medical perspective by making it viable to conduct wider studies on a larger patient base.