Multi-Scale Tree Classifier Based Support for Spirometry Diagnosis

Abstract

Background: The assessment of pulmonary function by Spirometry is essential in the diagnosis and discrimination of respiratory disorders. However, successful spirometric investigation by a medical doctor necessitates a non-negligible amount of data. Hence, there is a call for automated assessment of spirometric parameters in order to facilitate the task of the physician. The presented work aims at applying a reliable multi-scale structure of tree classifiers to distinguish between healthy/abnormal states, obstructive/restrictive diseases and their severity levels (mild, moderate, severe-to-very severe) via classification of certain pulmonary function parameters. The spirometric data processing and parameter measurement are conducted with the help of a LabVIEW interface. Results: The proposed method leads to a procedure with an average overall accuracy of 90.52%. Conclusion: The suggested method is a fast easy-to-use classification with a potential to enable understanding of the underlying physiological differences between classes.