gms | German Medical Science

33rd International Congress on Electrocardiology

International Society of Electrocardiology

Low Information Entropy of Heart Rate is Strongly Associated with Myocardial Infarction

Meeting Abstract

  • corresponding author presenting/speaker S. Lau - Friedrich-Schiller-University Jena, Jena, Germany
  • J. Haueisen - Technical University of Ilmenau, Ilmenau, Germany
  • E.G. Schukat-Talamazzini - Friedrich-Schiller-University Jena, Jena, Germany
  • M. Baumert - University of Adelaide, Adeaide, Australia
  • M. Goemig - Friedrich-Schiller-University Jena, Jena, Germany
  • U. Leder - Friedrich-Schiller-University Jena, Jena, Germany
  • H.R. Figulla - Friedrich-Schiller-University Jena, Jena, Germany
  • A. Voss - University of Applied Sciences, Jena, Germany

33rd International Congress on Electrocardiology. Cologne, 28.06.-01.07.2006. Düsseldorf, Köln: German Medical Science; 2007. Doc06ice130

The electronic version of this article is the complete one and can be found online at:

Published: February 8, 2007

© 2007 Lau et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.



Question: Heart rate variability (HRV) is a marker of autonomic activity in the heart. A decreased HRV after myocardial infarction (MI) indicates a high risk of sudden cardiac death. Our hypothesis is that the information entropy of the beat-to-beat intervals is an appropriate measure of HRV in cardiovascular diseases. We investigated the effect of MI on HRV entropy.

Method Used: Datasets of age-matched study groups of 55 controls (CON) and 59 patients after MI were recorded. The information entropy was estimated by compressing the quantized successive beat-to-beat interval differences using Burrows-Wheeler-compression (Bzip2). This estimate was divided by meanNN (mean of intervals) to yield the normalized Bzip2 entropy (NBE). NBE was compared with standard time and frequency HRV measures to separate CON from MI. Decision trees were generated based on standard measures (TREE1) and based on standards and NBE (TREE2) respectively.

Results: The CON-MI mean difference was most significant in NBE with a p-value of 0.00027 in a Wilcoxon test. The time domain measure separating best was cvNN = sdNN/meanNN with 0.0033 (sdNN = std. deviation of intervals). The frequency domain measure separating best was LF (power in low frequencies 0.04-0.15 Hz) with 0.0152. Multivariately, TREE1 (Figure 1 [Fig. 1]) achieved a classification rate of 65% during leave-one-out validation (HF = power in high frequencies 0.15-0.4 Hz). TREE2 was less complex, but had a higher classification rate of 70%. Thus, a simple entropy threshold separated CON and MI better than the best standards-based decision tree.

Conclusion: The strong association between MI and low HRV entropy confirms our hypothesis. The information entropy is an appropriate measure of HRV and might have a high potential for the assessment of cardiac mortality.