Abstract
In idiopathic pulmonary fibrosis (IPF) breathing pattern changes with disease progress. This study aims to determine if unsupervised hierarchal cluster analysis (HCA) can be used to define airflow profile differences in people with and without IPF. This was tested using 31 patients with IPF and 17 matched healthy controls, all of whom had their lung function assessed using spirometry and carbon monoxide CO transfer. A resting tidal breathing (RTB) trace of two minutes duration was collected at the same time. A Euclidian distance technique was used to perform HCA on the airflow data. Four distinct clusters were found, with the majority (18 of 21, 86%) of the severest IPF participants (Stage 2 and 3) being in two clusters. The participants in these clusters exhibited a distinct minute ventilation (p < 0.05), compared to the other two clusters. The respiratory drive was greatest in Cluster 1, which contained many of the IPF participants. Unstructured HCA was successful in recognising different airflow profiles, clustering according to differences in flow rather than time. HCA showed that there is an overlap in tidal airflow profiles between healthy RTB and those with IPF. The further application of HCA in recognising other respiratory disease is discussed.