Multi-Gaussian fitting for Digital Volume Pulse using Weighted Least Squares method

Abstract

Analysis of Digital Volume Pulse (DVP) waveform is a low cost non-invasive method of obtaining vital information related to condition of cardiovascular system. In resent years, different Pulse Decomposition Analysis (PDA) methods have been applied to DVP waveform analysis and clinical meaningful parameters have been obtained. All these methods decomposed single-cycle DVP into a fixed number (such as 2, 4 or 5) of individual pulses. In this paper, a Multi-Gaussian (MG) model is proposed to approximate DVP waveform using a variable number (4 or 5 in our study) of Gaussian pulses with normalized root MSE<2.0%. Unknown parameters in MG model are estimated by Weighted Least Squares (WLS) method. Performance of MG model and WLS are evaluated by fitting 80 DVP waveforms with four different types. High agreement between DVP waveform and MG model shows that the model can be used to analyse DVP waveforms.