Feasibility of Nonlinear Heart Rate Variability Analysis in Clinical Settings

The measure of heart rate variability (HRV) has become a valuable metric for diagnosing cardiac health. The ECG is the representative signal containing information about the condition of this health metric. Analysis of this highly complex and irregular signal cannot always be addressed through linear statistics. Nonlinear methods are able to describe the processes generated by biological systems in a more effective way. The adoption of these methods in a clinical environment, however, has been difficult and slow. This paper examines the feasibility of using nonlinear analysis methods in such a setting. Given two data sets of a normal patient and a patient with atrial fibrillation (from PhysioNet), we examined the effectiveness of using Poincaré plots, largest Lyapunov exponent, and detrended fluctuation analysis, in differentiating the subjects. All the methods used were able to clearly separate the two data sets. From a clinical perspective, calculating accurate Lyapunov exponents requires an average of 5.5 hours of data, while Poincaré plots and DFA require approximately 100 and 80 minutes, respectively. Both Poincaré plots and DFA would serve well in characterizing a patient relatively quickly, while Lyapunov exponents would be too time intensive. To test our hypothesis, we designed and implemented a simple ECG system that gathered 90 minutes of data from an unclassified subject. A Poincaré and DFA analysis of the data suggested a healthy normal individual.