Analysis of cardiopulmonary dynamics using wavelet analysis methods: removing the influence of respiration and prediction of emphysema

Respiratory effects are significant in standard frequency analysis of heart rate variability (HRV). This limits evaluation of sympathovagal balance in exercise and other dynamic states. We developed a method using wavelet statistical analyses combined with cluster analysis to reduce this. Methods: Ten emphysema subjects (ES) and 5 matched controls (CS) were evaluated using wavelet entropy (WE) of heart rate data during rest and exercise. The model included wavelet source separation via a discrete wavelet decomposition of heart rate and respiration signals. Data were analyzed blinded, and prediction for group separation (ES versus CS) was performed using cluster analysis. Results: Using WE, the source separation model was able to predict actual classification of individuals with 93% accuracy at rest (4/5 CS, 10/10 ES) and 100% accuracy at exercise. The most significant factors for principal component analysis were identified for resting and exercise states. Conclusion: Novel utilization of wavelet analysis in combination with principal component analysis and wavelet analysis methods can remove the influences of respiration on the time frequency analysis of HRV. Support: VIDDA Foundation