Quantitative delineation of heart murmurs using features derived from autoregressive modeling

In this paper, the focus is on systolic heart murmurs of clinical significance, quantitative features characterizing the murmurs are derived by dividing the systole into many short non-overlapping segments and using second order autoregressive (AR) models. Features thus derived can provide a quantitative delineation of the murmur with respect to the onset, duration, intensity and pitch (frequency). We applied the said approach to examine several systolic murmurs and obtained rather accurate descriptions that parallel closely to published clinical documentation.