Determination of Respiratory Parameters by Means of Hurst Exponents of the Respiratory Sounds and Stochastic Processing Methods

Abstract

Objectives: System approach to the human respiratory system and input/output signals which characterize the system properties were not explored in detail in the literature. The aim of this study is to propose a combination of methods to investigate the indirect relationship between the fractal properties of Respiratory Signals (RS) and Respiratory Sound Signals (RSS) and the clinically measured respiratory parameters. Methods: We used Hurst exponent to reveal the fractal properties of RS and RSS and to estimate the pressures in the respiratory system. The combination of well-known statistical signal processing methods and optimization were applied to the experimentally acquired 23 records. Pearson correlation coefficient and Bland-Altman analysis were the chosen validation methods. Results: Considerable amounts of Hurst exponent values of RSS were found to be between 0.5 and 1, which means increasing trend or decreasing trend can be seen in RSS with fractional Gaussian process properties. Results of the pressure estimator revealed that internal pressure due to tissue viscoelasticity is higher than the pressure due to static elasticity. Feature power and skewness also provided distinctive results for all recordings. Conclusion: Hurst exponent values of the RSS are fruitful representation of the signals which bring the underlaying system characteristics into the surface. We illustrated that required number of sensors can be reduced in the feature calculation to ease implementation effort on the hardware of the handheld devices. Significance: Bland-Altman plots were very successful to demonstrate the connection between the sets of measured respiratory parameters and calculated features.