Detection Performance and Risk Stratification Using a Model-Based Shape Index Characterizing Heart Rate Turbulence
Author
Summary, in English
A detection-theoretic approach to quantify heart rate turbulence (HRT) following a ventricular premature beat is proposed and validated using an extended integral pulse frequency modulation (IPFM) model which accounts for HRT. The modulating signal of the extended IPFM model is projected into a three-dimensional subspace spanned by the Karhunen-Loeve basis functions, characterizing HRT shape. The presence or absence of HRT is decided by means of a likelihood ratio test, the Neyman-Pearson detector, resulting in a quadratic detection statistic. Using a labeled dataset built from different interbeat interval series, detection performance is assessed and found to outperform the two widely used indices: turbulence onset (TO) and turbulence slope (TS). The ability of the proposed method to predict the risk of cardiac death is evaluated in a population of patients (n = 90) with ischemic cardiomyopathy and mild-to-moderate congestive heart failure. While both TS and the novel HRT index differ significantly in survivors and cardiac death patients, mortality analysis shows that the latter index exhibits much stronger association with risk of cardiac death (hazard ratio = 2.8, CI = 1.32-5.97, p = 0.008). It is also shown that the model-based shape indices, but not TO and TS, remain predictive of cardiac death in our population when computed from 4-h instead of 24-h ambulatory ECGs.
Publishing year
2010
Language
English
Pages
3173-3184
Publication/Series
Annals of Biomedical Engineering
Volume
38
Issue
10
Document type
Journal article
Publisher
Springer
Topic
- Electrical Engineering, Electronic Engineering, Information Engineering
Keywords
- stratification
- Risk
- Mortality analysis
- Detection theory
- Karhunen-Loeve transform
- Likelihood ratio test
- Heart rate turbulence
- Neyman-Pearson detection
- Congestive heart failure
- Ischemic cardiomyopathy
Status
Published
Research group
- Signal Processing
ISBN/ISSN/Other
- ISSN: 1573-9686