Independent component analysis based source separation in physiological strain waves

Physiological cyclic strain waves in tissue are caused by the pulsation of blood flow in arterial vasculature due to cardiac activity and modulation of blood volume in venous vasculature due to respiration. We applied independent component analysis ( ICA) to extract the sources of the strain wave modeled as three linear mixtures: 1) cardiac source; 2) respiratory source; 3) other sources. Factor analysis (FA) with minimum description length (MDL) was applied at the preprocessing step to determine the number of sources. To test the efficacy of FA and MDL, we implemented two hundred iterations of simulated source separations on model signals formed from summing simulated cardiac and respiratory sources. We obtained the correct source number for 176 out of 200 iterations. The extracted sources matched the original sources with correlation coefficients higher than 95% in successive field simulations. We also applied the ICA strategy to an in-vivo study and extracted three sources which included cardiac and respiratory sources. The results demonstrate the efficacy of ICA in strain wave source separation.