Lung sound extraction from mixed lung and heart sounds FASTICA algorithm

Abstract

The reduction of interface by heart sounds in lung sounds, in order to improve the analysis of lung sounds has been discussed since the earliest work in 1986 [1] and numerous numbers of papers were developed further. In fact, a large amount of methods was proposed to quantify pure lung sounds, compared with heart sounds interference [2-5]. This paper deals with the performance of lung and heart sound separation based on an objective method, which consist on a mathematical independent component analysis algorithm applied to a preprocessed waveform and spectra of lung sounds recorded with a digital stethoscope. However, such objective evaluation allows to interpret the significance of the noise reduction and to conclusively compare the performance of this particular method with other methods. The quantitative performance measure of those techniques is almost based on the NR ratio or NIR ratio, which is the noise to input ratio, in some references and also the S/N ratio that is the signal to noise ratio. In this investigation we will introduce the bases of the ICA technique used to lung and heart signal separation, we will expose some results of lung and heart signal separation that we have obtained based on the methodology described above. This method was successfully used to separate heart sound and lung sound from recorded mixed sounds. A number of illustrations are exposed here and a heart to lung sound signal ratio H/L is calculated for different cases.