Cross-recurrence analysis for pattern matching of multidimensional physiological signals.

Cross-recurrence quantification analysis (CRQA), based on the cross-recurrence plot (CRP), is an effective method to characterize and quantify the nonlinear interrelationships between a pair of nonlinear time series. It allows the flexibility of reconstructing signals in the phase space and to identify different types of patterns at arbitrary positions between trajectories. These advantages make CRQA attractive for time series data mining tasks, which have been of recent interest in the literature. However, little has been done to exploit CRQA for pattern matching of multidimensional, especially spatiotemporal, physiological signals. In this paper, we present a novel methodology in which CRQA statistics serve as measures of dissimilarity between pairs of signals and are subsequently used to uncover clusters within the data. This methodology is evaluated on a real dataset consisting of 3D spatiotemporal vectorcardiogram (VCG) signals from healthy and diseased patients. Experimental results show that Lmax, the length of the longest diagonal line in the CRP, yields the best-performing clustering that almost exactly matches the ground truth diagnoses of patients. Results also show that our proposed measure, Rτ max, which characterizes the maximum similarity between signals over all pairwise time-delayed alignments, outperforms all other tested CRQA measures (in terms of matching the ground truth) when the VCG signals are rescaled to reduce the effects of signal amplitude.

[1]  Basabi Chakraborty,et al.  Effect of Data Representation for Time Series Classification - A Comparative Study and a New Proposal , 2019, Mach. Learn. Knowl. Extr..

[2]  Hui Yang,et al.  Spatiotemporal Differentiation of Myocardial Infarctions , 2013, IEEE Transactions on Automation Science and Engineering.

[3]  Tak-Chung Fu,et al.  A review on time series data mining , 2011, Eng. Appl. Artif. Intell..

[4]  A. Giuliani,et al.  Detecting deterministic signals in exceptionally noisy environments using cross-recurrence quantification , 1998 .

[5]  Derek Abbott,et al.  Beat-to-Beat Vectorcardiographic Analysis of Ventricular Depolarization and Repolarization in Myocardial Infarction , 2012, PloS one.

[6]  A. Giuliani,et al.  Using woody genera for phytogeographic regionalization at a medium scale: A case study of Italy , 2016 .

[7]  Nima Hatami,et al.  Classification of time-series images using deep convolutional neural networks , 2017, International Conference on Machine Vision.

[8]  R. Andrzejak,et al.  Cross recurrence quantification for cover song identification , 2009 .

[9]  Jürgen Kurths,et al.  Recurrence plots for the analysis of complex systems , 2009 .

[10]  Hui Yang,et al.  Multiscale Recurrence Quantification Analysis of Spatial Cardiac Vectorcardiogram Signals , 2011, IEEE Transactions on Biomedical Engineering.

[11]  T. Chou,et al.  When is the vectorcardiogram superior to the scalar electrocardiogram? , 1986, Journal of the American College of Cardiology.

[12]  M. Thiel,et al.  Cross recurrence plot based synchronization of time series , 2002, physics/0201062.

[13]  K Lyttkens,et al.  Vectorcardiogram more sensitive than 12-lead ECG in the detection of inferior myocardial infarction. , 1990, Clinical physiology.

[14]  Juan Pablo Bello,et al.  Measuring Structural Similarity in Music , 2011, IEEE Transactions on Audio, Speech, and Language Processing.

[15]  O Pahlm,et al.  Increased sensitivity for the diagnosis of healed myocardial infarction using vectorial information in the 12-lead ECG. , 1995, Journal of electrocardiology.

[16]  Celso Ferreira Filho,et al.  Significance of Vectorcardiogram in the Cardiological Diagnosis of the 21st Century , 2007, Clinical cardiology.

[17]  N. Marwan,et al.  Nonlinear analysis of bivariate data with cross recurrence plots , 2002, physics/0201061.

[18]  Radek Martinek,et al.  Comparison of Different Electrocardiography with Vectorcardiography Transformations , 2019, Sensors.

[19]  Tania Ghosal,et al.  Planarity of the spatial QRS loop of vectorcardiogram is a crucial diagnostic and prognostic parameter in acute myocardial infarction. , 2019, Medical hypotheses.

[20]  Hui Yang,et al.  Self-organized topology of recurrence-based complex networks. , 2013, Chaos.

[21]  Satish T. S. Bukkapatnam,et al.  Topology and Random-Walk Network Representation of Cardiac Dynamics for Localization of Myocardial Infarction , 2013, IEEE Transactions on Biomedical Engineering.

[22]  Jonathan Scott Friedlaender,et al.  Partial correlation of distance matrices in studies of population structure. , 1987, American journal of physical anthropology.

[23]  Hui Yang,et al.  Spatiotemporal representation of cardiac vectorcardiogram (VCG) signals , 2012, Biomedical engineering online.

[24]  Eleni I. Vlahogianni,et al.  Temporal Evolution of Short‐Term Urban Traffic Flow: A Nonlinear Dynamics Approach , 2008, Comput. Aided Civ. Infrastructure Eng..

[25]  Georg Zitzlsberger,et al.  Highly scalable algorithm for computation of recurrence quantitative analysis , 2018, The Journal of Supercomputing.

[26]  Hui Yang,et al.  Multiscale Adaptive Basis Function Modeling of Spatiotemporal Vectorcardiogram Signals , 2013, IEEE Journal of Biomedical and Health Informatics.

[27]  David Schultz,et al.  Approximation of diagonal line based measures in recurrence quantification analysis , 2015 .

[28]  S. Dandapat,et al.  Detection of myocardial infarction from vectorcardiogram using relevance vector machine , 2017, Signal Image Video Process..

[29]  Hui Yang,et al.  Linear affine transformations between 3-lead (Frank XYZ leads) vectorcardiogram and 12-lead electrocardiogram signals. , 2009, Journal of electrocardiology.

[30]  Hui Yang,et al.  Self-organizing visualization and pattern matching of vectorcardiographic QRS waveforms , 2016, Comput. Biol. Medicine.

[31]  Charles L. Webber,et al.  Cross recurrence quantification of coupled oscillators , 2002 .

[32]  Moreno I. Coco,et al.  Cross-recurrence quantification analysis of categorical and continuous time series: an R package , 2013, Front. Psychol..

[33]  Hui Yang,et al.  Identification of myocardial infarction (MI) using spatio-temporal heart dynamics. , 2012, Medical engineering & physics.