George B. Moody PhysioNet Challenge

Logo

Quick links for this year's Challenge:

Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address.

Session S41.1

A Method For Characterising The Human RR Tachogram To Identify Normal Circadian Variation

G.D. Clifford, P.E. McSharry

University of Oxford

Oxford, UK

A typical 24-hour RR tachogram derived from the electrocardiogram of a normal healthy human is characterised by a few dominant frequency components which, together with their amplitudes and the local mean RR interval, typically change slowly over a period of minutes. Within such segments, cardiovascular activity is assumed to be almost stationary and frequency domain analysis is often employed. A shift in cardiovascular state is associated with a change in both the local mean RR interval and the relative contributions of the component frequencies. Such state switching is often accompanied by changes in physical or mental activity and therefore may be an indicator of the circadian activity of a subject.

Recent clinical studies have indicated that aspects of circadian activity in humans are linked to long term health indicators such as cardiovascular metrics. The authors present methods for describing and quantifying these circadian changes as an aid to detecting normal human cardiovascular activity (and therefore deviations from normality).

Furthermore, the incidences of ectopy and artefact as a function of time of day, heart rate and state changes for normal subjects are investigated. Results show that artefact is significantly correlated with state change and heart rate in normal humans, whereas ectopy exhibits no significant relationship, with an ectopic beat occurring, on average, once an hour. Artefact is therefore shown to be a source of information which can aid identification of state changes and facilitates abnormality detection.

A scoring system is proposed that correctly identifies a significant proportion of the real (normal) and artificial RR interval time series in event 2 of the Physionet/Computers in Cardiology Challenge 2002 (entry number 12). This paper forms the basis of the reasoning behind an accompanying paper ‘A Method for generating synthetic RR tachograms of normal humans over 24-hours’ and entry number 201 in event 1 of the Challenge.

Session S51.1

A Method For Generating Synthetic RR Tachograms Of Normal Humans Over 24-Hours

P.E. McSharry, G.D. Clifford

University of Oxford

Oxford, UK

An algorithm that generates realistic synthetic 24-hour RR tachograms by including both cardiovascular interactions and transitions between physiological states is presented. Fluctuations of the beat to beat RR intervals of a normal healthy human over 24-hours are known to exhibit variability on a number of different time scales. Short range variability due to Mayer waves and RSA are incorporated into the algorithm using a power spectrum with given spectral characteristics described by its low frequency and high frequency components respectively. Longer range fluctuations arising from transitions between physiological states are generated using switching distributions extracted from real data. These physiological states, including sleep states, are specified using RR intervals with particular means and trends. An analysis of ectopic beat and artefact incidence in an accompanying paper (“A method for characterising the human RR tachogram to identify normal circadian variation”) is used to provide a mechanism for generating realistic ectopy and artefact. Ectopic beats are added with an independent probability of one per hour. Artefacts are included with a probability proportional to mean heart rate within a state and increased for state transition periods. This algorithm provides RR tachograms that are similar to those in the MIT-BIH Normal Sinus Rhythm Database. The resulting synthetic RR interval generator has been submitted to part 1 of the Physionet/Computers in Cardiology Challenge 2002 with entry number 201.


Supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under NIH grant number R01EB030362.

Back