Clustering analysis of respiratory crackles

Abstract

Crackles extracted from lung sounds recorded from pathological subjects were modelled using two different sets of parameters. The first set consisted of conventional parameters used widely by physicians. The second set consisted of Prony frequencies and damping coefficients obtained from the Prony model of order one. The crackles were clustered with no a priori information about the diseases using iterative partitional clustering. Clustering using either models shows that there is a relationship between crackles and lung diseases.