Causality in physiological signals

Andreas Mueller, Jan F. Kraemer, Thomas Penzel, Hendrik Bonnemeier, Juergen Kurths, Niels Wessel*

*Corresponding author for this work

Research output: Contribution to journalLiterature review

18 Citations (Scopus)

Abstract

Health is one of the most important non-material assets and thus also has an enormous influence on material values, since treating and preventing diseases is expensive. The number one cause of death worldwide today originates in cardiovascular diseases. For these reasons the aim of understanding the functions and the interactions of the cardiovascular system is and has been a major research topic throughout various disciplines for more than a hundred years. The purpose of most of today's research is to get as much information as possible with the lowest possible effort and the least discomfort for the subject or patient, e.g. via non-invasive measurements. A family of tools whose importance has been growing during the last years is known under the headline of coupling measures. The rationale for this kind of analysis is to identify the structure of interactions in a system of multiple components. Important information lies for example in the coupling direction, the coupling strength, and occurring time lags. In this work, we will, after a brief general introduction covering the development of cardiovascular time series analysis, introduce, explain and review some of the most important coupling measures and classify them according to their origin and capabilities in the light of physiological analyses. We will begin with classical correlation measures, go via Granger-causality-based tools, entropy-based techniques (e.g. momentary information transfer), nonlinear prediction measures (e.g. mutual prediction) to symbolic dynamics (e.g. symbolic coupling traces). All these methods have contributed important insights into physiological interactions like cardiorespiratory coupling, neuro-cardio-coupling and many more. Furthermore, we will cover tools to detect and analyze synchronization and coordination (e.g. synchrogram and coordigram). As a last point we will address time dependent couplings as identified using a recent approach employing ensembles of time series. The scope of this review, as opposed to various other excellent reviews like (Hlavackova-Schindler et al Phys. Rep. 441 1-46, Kramer et al 2004 Phys. Rev. E 70 1-10, Lombardi 2000 Circulation 101 8-10, Porta et al 2000 Am. J. Physiol.: Heart and Circulatory Physiol. 279 H2558-67, Schelter et al 2006 J. Neurosci. Methods 152 210-9), is to give a broader overview over existing coupling measures and where to look to find the most appropriate tool for a given situation. The review will comprise a test of one representative of the most important coupling measure groups using a simple toy model to illustrate some essential features of the tools. At the end we will summarise the performance of each measure and offer some advice on when to use which method.

Original languageEnglish
Pages (from-to)R46-R72
Number of pages27
JournalPhysiological Measurement
Volume37
Issue number5
DOIs
Publication statusPublished - 21 Apr 2016

Keywords

  • coupling direction
  • time series analysis
  • cardiovascular system
  • heart-rate-variability
  • partial directed coherence
  • obstructive sleep-apnea
  • power spectrum analysis
  • bivariate time-series
  • Granger causality
  • cardiovascular variability
  • nonlinear dynamics
  • blood-pressure
  • cardiorespiratory interaction

Cite this

Mueller, A., Kraemer, J. F., Penzel, T., Bonnemeier, H., Kurths, J., & Wessel, N. (2016). Causality in physiological signals. Physiological Measurement, 37(5), R46-R72. https://doi.org/10.1088/0967-3334/37/5/R46

Causality in physiological signals. / Mueller, Andreas; Kraemer, Jan F.; Penzel, Thomas; Bonnemeier, Hendrik; Kurths, Juergen; Wessel, Niels.

In: Physiological Measurement, Vol. 37, No. 5, 21.04.2016, p. R46-R72.

Research output: Contribution to journalLiterature review

Mueller, A, Kraemer, JF, Penzel, T, Bonnemeier, H, Kurths, J & Wessel, N 2016, 'Causality in physiological signals', Physiological Measurement, vol. 37, no. 5, pp. R46-R72. https://doi.org/10.1088/0967-3334/37/5/R46
Mueller A, Kraemer JF, Penzel T, Bonnemeier H, Kurths J, Wessel N. Causality in physiological signals. Physiological Measurement. 2016 Apr 21;37(5):R46-R72. https://doi.org/10.1088/0967-3334/37/5/R46
Mueller, Andreas ; Kraemer, Jan F. ; Penzel, Thomas ; Bonnemeier, Hendrik ; Kurths, Juergen ; Wessel, Niels. / Causality in physiological signals. In: Physiological Measurement. 2016 ; Vol. 37, No. 5. pp. R46-R72.
@article{5f5894248cca4b8ebf301f6b7aa28e7f,
title = "Causality in physiological signals",
abstract = "Health is one of the most important non-material assets and thus also has an enormous influence on material values, since treating and preventing diseases is expensive. The number one cause of death worldwide today originates in cardiovascular diseases. For these reasons the aim of understanding the functions and the interactions of the cardiovascular system is and has been a major research topic throughout various disciplines for more than a hundred years. The purpose of most of today's research is to get as much information as possible with the lowest possible effort and the least discomfort for the subject or patient, e.g. via non-invasive measurements. A family of tools whose importance has been growing during the last years is known under the headline of coupling measures. The rationale for this kind of analysis is to identify the structure of interactions in a system of multiple components. Important information lies for example in the coupling direction, the coupling strength, and occurring time lags. In this work, we will, after a brief general introduction covering the development of cardiovascular time series analysis, introduce, explain and review some of the most important coupling measures and classify them according to their origin and capabilities in the light of physiological analyses. We will begin with classical correlation measures, go via Granger-causality-based tools, entropy-based techniques (e.g. momentary information transfer), nonlinear prediction measures (e.g. mutual prediction) to symbolic dynamics (e.g. symbolic coupling traces). All these methods have contributed important insights into physiological interactions like cardiorespiratory coupling, neuro-cardio-coupling and many more. Furthermore, we will cover tools to detect and analyze synchronization and coordination (e.g. synchrogram and coordigram). As a last point we will address time dependent couplings as identified using a recent approach employing ensembles of time series. The scope of this review, as opposed to various other excellent reviews like (Hlavackova-Schindler et al Phys. Rep. 441 1-46, Kramer et al 2004 Phys. Rev. E 70 1-10, Lombardi 2000 Circulation 101 8-10, Porta et al 2000 Am. J. Physiol.: Heart and Circulatory Physiol. 279 H2558-67, Schelter et al 2006 J. Neurosci. Methods 152 210-9), is to give a broader overview over existing coupling measures and where to look to find the most appropriate tool for a given situation. The review will comprise a test of one representative of the most important coupling measure groups using a simple toy model to illustrate some essential features of the tools. At the end we will summarise the performance of each measure and offer some advice on when to use which method.",
keywords = "coupling direction, time series analysis, cardiovascular system, heart-rate-variability, partial directed coherence, obstructive sleep-apnea, power spectrum analysis, bivariate time-series, Granger causality, cardiovascular variability, nonlinear dynamics, blood-pressure, cardiorespiratory interaction",
author = "Andreas Mueller and Kraemer, {Jan F.} and Thomas Penzel and Hendrik Bonnemeier and Juergen Kurths and Niels Wessel",
note = "Acknowledgment TP was supported by project no. LQ1605 from the National Program of Sustainability II and FNUSA-ICRC (No. Z.1.05/1.1.00/02.0123). AM and NW were supported by the German Research Foundation (WE 2835/5-1).",
year = "2016",
month = "4",
day = "21",
doi = "10.1088/0967-3334/37/5/R46",
language = "English",
volume = "37",
pages = "R46--R72",
journal = "Physiological Measurement",
issn = "0967-3334",
publisher = "IOP Publishing Ltd.",
number = "5",

}

TY - JOUR

T1 - Causality in physiological signals

AU - Mueller, Andreas

AU - Kraemer, Jan F.

AU - Penzel, Thomas

AU - Bonnemeier, Hendrik

AU - Kurths, Juergen

AU - Wessel, Niels

N1 - Acknowledgment TP was supported by project no. LQ1605 from the National Program of Sustainability II and FNUSA-ICRC (No. Z.1.05/1.1.00/02.0123). AM and NW were supported by the German Research Foundation (WE 2835/5-1).

PY - 2016/4/21

Y1 - 2016/4/21

N2 - Health is one of the most important non-material assets and thus also has an enormous influence on material values, since treating and preventing diseases is expensive. The number one cause of death worldwide today originates in cardiovascular diseases. For these reasons the aim of understanding the functions and the interactions of the cardiovascular system is and has been a major research topic throughout various disciplines for more than a hundred years. The purpose of most of today's research is to get as much information as possible with the lowest possible effort and the least discomfort for the subject or patient, e.g. via non-invasive measurements. A family of tools whose importance has been growing during the last years is known under the headline of coupling measures. The rationale for this kind of analysis is to identify the structure of interactions in a system of multiple components. Important information lies for example in the coupling direction, the coupling strength, and occurring time lags. In this work, we will, after a brief general introduction covering the development of cardiovascular time series analysis, introduce, explain and review some of the most important coupling measures and classify them according to their origin and capabilities in the light of physiological analyses. We will begin with classical correlation measures, go via Granger-causality-based tools, entropy-based techniques (e.g. momentary information transfer), nonlinear prediction measures (e.g. mutual prediction) to symbolic dynamics (e.g. symbolic coupling traces). All these methods have contributed important insights into physiological interactions like cardiorespiratory coupling, neuro-cardio-coupling and many more. Furthermore, we will cover tools to detect and analyze synchronization and coordination (e.g. synchrogram and coordigram). As a last point we will address time dependent couplings as identified using a recent approach employing ensembles of time series. The scope of this review, as opposed to various other excellent reviews like (Hlavackova-Schindler et al Phys. Rep. 441 1-46, Kramer et al 2004 Phys. Rev. E 70 1-10, Lombardi 2000 Circulation 101 8-10, Porta et al 2000 Am. J. Physiol.: Heart and Circulatory Physiol. 279 H2558-67, Schelter et al 2006 J. Neurosci. Methods 152 210-9), is to give a broader overview over existing coupling measures and where to look to find the most appropriate tool for a given situation. The review will comprise a test of one representative of the most important coupling measure groups using a simple toy model to illustrate some essential features of the tools. At the end we will summarise the performance of each measure and offer some advice on when to use which method.

AB - Health is one of the most important non-material assets and thus also has an enormous influence on material values, since treating and preventing diseases is expensive. The number one cause of death worldwide today originates in cardiovascular diseases. For these reasons the aim of understanding the functions and the interactions of the cardiovascular system is and has been a major research topic throughout various disciplines for more than a hundred years. The purpose of most of today's research is to get as much information as possible with the lowest possible effort and the least discomfort for the subject or patient, e.g. via non-invasive measurements. A family of tools whose importance has been growing during the last years is known under the headline of coupling measures. The rationale for this kind of analysis is to identify the structure of interactions in a system of multiple components. Important information lies for example in the coupling direction, the coupling strength, and occurring time lags. In this work, we will, after a brief general introduction covering the development of cardiovascular time series analysis, introduce, explain and review some of the most important coupling measures and classify them according to their origin and capabilities in the light of physiological analyses. We will begin with classical correlation measures, go via Granger-causality-based tools, entropy-based techniques (e.g. momentary information transfer), nonlinear prediction measures (e.g. mutual prediction) to symbolic dynamics (e.g. symbolic coupling traces). All these methods have contributed important insights into physiological interactions like cardiorespiratory coupling, neuro-cardio-coupling and many more. Furthermore, we will cover tools to detect and analyze synchronization and coordination (e.g. synchrogram and coordigram). As a last point we will address time dependent couplings as identified using a recent approach employing ensembles of time series. The scope of this review, as opposed to various other excellent reviews like (Hlavackova-Schindler et al Phys. Rep. 441 1-46, Kramer et al 2004 Phys. Rev. E 70 1-10, Lombardi 2000 Circulation 101 8-10, Porta et al 2000 Am. J. Physiol.: Heart and Circulatory Physiol. 279 H2558-67, Schelter et al 2006 J. Neurosci. Methods 152 210-9), is to give a broader overview over existing coupling measures and where to look to find the most appropriate tool for a given situation. The review will comprise a test of one representative of the most important coupling measure groups using a simple toy model to illustrate some essential features of the tools. At the end we will summarise the performance of each measure and offer some advice on when to use which method.

KW - coupling direction

KW - time series analysis

KW - cardiovascular system

KW - heart-rate-variability

KW - partial directed coherence

KW - obstructive sleep-apnea

KW - power spectrum analysis

KW - bivariate time-series

KW - Granger causality

KW - cardiovascular variability

KW - nonlinear dynamics

KW - blood-pressure

KW - cardiorespiratory interaction

U2 - 10.1088/0967-3334/37/5/R46

DO - 10.1088/0967-3334/37/5/R46

M3 - Literature review

VL - 37

SP - R46-R72

JO - Physiological Measurement

JF - Physiological Measurement

SN - 0967-3334

IS - 5

ER -