Pulse Wave Analysis by Quantified Reconstructed Attractors

Abstract

Direct time domain analysis is a classical method for pulse wave analysis. Attractor reconstruction is a new and alternative way to analyse the morphology of a pulse wave. The novel approach plots the pulse wave data as overlapping loops in the three-dimensional phase space and, subsequently, projects the trajectories onto a plane generating two-dimensional attractors. This paper presents methods that automatically quantify these attractors. At first, a short overview of the attractor reconstruction technique is provided. After preprocessing of the pulse waveform data, the attractor is generated and edited by a median filter to remove remaining artefacts that can distort the quantification process. The developed algorithm is based on image processing and extracts attractor features such as the angle of rotation, the angle between the arm fragments, shape, height, lengths and width of the attractor. Finally, visual examination evaluates the performance of the feature extraction technique revealing that the majority of the generated attractors are correctly quantified. However, it also shows that remaining outliers and metrological discrepancies have a negative impact on the results.