Polynomial regression of lung sound AR parameters on flow and volume

Earlier studies of our group have shown that autoregressive (AR) parameters of lung sounds clearly depend on airflow and lungvolume. In this report, it is shown that 3rd degree polynomials in flow and volume can be used to describe AR models of lung sounds. These polynomial AR models produce prediction errors comparable to those of AR models fitted to the raw data. Since the polynomials are described by a relatively small number of caefficients, they allow a considerable datareduction. Moreover, it is shown that the polynomials are influenced by induced airway narrowing. This effect might eventually constitute a tool for the diagnosis of airway obstructing diseases, and other lung diseases that affect the lung sounds.