The Sounds of Silence

Background: Phonocardiography is a non-invasive method to diagnose human heart condition. The contraction and relaxation of the heart muscle, and the movement of the heart valves generate audible vibrations, which can be evaluated by a qualified cardiologist. However, some of the information that we look for exists at the infrasound area of the frequency domain. Owing to the development of the CompuStethTM electronic stethoscope of Bat-Call it is now possible to analyze the heart audible sounds with their infrasound. The purpose of this study was to classify normal and aortic stenosis heart sounds and to show the added value of the infrasound data to the classification.

Methods: 1,399 heart sounds were collected and classified. The dataset contained 629 normal sounds and 770 aortic stenosis sounds. Different kinds of classifiers were applied to the data, and the specificity, sensitivity and accuracy were calculated.

Results: The results of the best classifier, while taking into consideration the infrasound, were: accuracy of 82%, sensitivity 83% and specificity 81%. When removing the low frequencies from the analysis (<60 Hz) the results were: accuracy of 75.6%, sensitivity 79% and specificity 72%.

Conclusion: Even in the case of aortic stenosis, which is a relatively easy case for diagnosis using only audible sounds, Infrasound has added value in the classification process. The potential of using Infrasound should be further investigated in the diagnosis of more challenging heart diseases.