Non-invasive infrasound heart murmur detection with a Support Vector Machine (SVM) classification approach

The goal of this paper is to present new devices and techniques that detect low frequency vibrations from the human chest and correlate them to cardiac conditions. Several new devices and techniques of detecting a human heart murmur have been developed through the extraction of vibrations primarily in the range of 10-150 Hertz (Hz) on the human chest. The devices and techniques have been tested on different types of simulators and through clinical trials with the consent of the University of Minnesota Institutional Review Board (IRB). Signals were collected using a Kardiac Infrasound Device (KID) and accelerometers integrated with a custom MATLAB software interface and a data acquisition system. Using the interface, the data was analyzed and classified by a Support Vector Machine (SVM) approach. Results show that the SVM was able to classify signals under different testing environments. For clinical trials, the SVM distinguished between normal and abnormal cardiac conditions and between pathological and non-pathological cardiac conditions. Finally, using the various devices, a correlation between heart murmurs and normal hearts was observed from human chest vibrations.