Time-frequency analysis and diagnosis of native heart valve sounds

Investigates the performance of the Choi-Williams time-frequency distribution as a technique for the analysis of heart sounds to determine the condition of aortic and mitral native heart valves. The time-frequency distributions of a total of 45 subjects were generated and 9 significant time and frequency features extracted from each. Ten of the subjects had normal aortic valves, 13 had dysfunctioning aortic valves, 10 had normal mitral valves and 12 had dysfunctioning mitral valves. From the collective results, a definition of normality was specified far each measured parameter in the aortic and mitral case. The performance of each subject with respect to these normality criteria was calculated as a total of conditions met and conditions failed. Using this count as a discriminative parameter, it is shown that a value of 98% correct classification is achieved. Results presented in this work clearly show that time-frequency spectral estimation is the correct tool for the analysis and assessment of heart sounds.