Investigation of nonuniform airflow signal oscillation during high frequency chest compression
 Kiwon Sohn^{1},
 Warren J Warwick^{2}Email author,
 Yong W Lee^{1},
 Jongwon Lee^{1} and
 James E Holte^{1}
DOI: 10.1186/1475925X434
© Sohn et al; licensee BioMed Central Ltd. 2005
Received: 10 March 2005
Accepted: 19 May 2005
Published: 19 May 2005
Abstract
Background
High frequency chest compression (HFCC) is a useful and popular therapy for clearing bronchial airways of excessive or thicker mucus. Our observation of respiratory airflow of a subject during use of HFCC showed the airflow oscillation by HFCC was strongly influenced by the nonlinearity of the respiratory system. We used a computational modelbased approach to analyse the respiratory airflow during use of HFCC.
Methods
The computational model, which is based on previous physiological studies and represented by an electrical circuit analogue, was used for simulation of in vivo protocol that shows the nonlinearity of the respiratory system. Besides, airflow was measured during use of HFCC. We compared the simulation results to either the measured data or the previous research, to understand and explain the observations.
Results and discussion
We could observe two important phenomena during respiration pertaining to the airflow signal oscillation generated by HFCC. The amplitudes of HFCC airflow signals varied depending on spontaneous airflow signals. We used the simulation results to investigate how the nonlinearity of airway resistance, lung capacitance, and inertance of air characterized the respiratory airflow. The simulation results indicated that lung capacitance or the inertance of air is also not a factor in the nonuniformity of HFCC airflow signals. Although not perfect, our circuit analogue model allows us to effectively simulate the nonlinear characteristics of the respiratory system.
Conclusion
We found that the amplitudes of HFCC airflow signals behave as a function of spontaneous airflow signals. This is due to the nonlinearity of the respiratory system, particularly variations in airway resistance.
Background
High Frequency Chest Compression (HFCC) [6, 9, 19, 29] is a useful and popular therapy for clearing bronchial airways of excessive or thick mucus since it does not require patients to do any directed efforts for respiration while therapy is given, unlike other airway clearance techniques such as active cycle of breathing or autogenic drainage [17]. A HFCC machine pumps air into an inflatable jacket worn by patient. By means of a surrogate piston, sine waveform compression pulses with frequencies ranging from ~5 Hz to ~21 Hz are supplied to the thorax of a patient through the jacket. These pulses squeeze and vibrate the patient's thorax at prescribed frequencies. These actions help in the evacuation of mucus through changing the rheological property of the mucus and airflow oscillation. King et al. [13], Krumpe et al. [15], and Tomkiewicz at al. [28], showed that HFCC pulses decrease the viscosity of mucus and helps evacuation. A more important consequence is the respiratory airflow oscillated during use of HFCC, which results from variation of intrapleural pressure. Lapin [17] and Warwick [30] pointed out that airflow is the most important factor for mucus transport since airflow produces the shear stresses for evacuation of mucus.
Although several models have been developed to simulate respiration by other researchers and their models successfully worked for their own purposes, these models are not appropriate for HFCC simulation because these models were either linearized [7, 8] or because they assumed that respiration was driven by a mechanical ventilator at the mouth [5, 20]. The model presented in this paper provides reliable simulation results on fast change of intrapleural pressure (P_{ pl }) altered by HFCC because this model is described with nonlinear equations and P_{ pl }is selected as the driving force of respiration. The nonlinear characteristics of the respiratory system are not easily noticed during quiet tidal breathing, but the airflow signal oscillation measured at the mouth during use of HFCC is a strong indicator of the nonlinear characteristics of the respiratory system.
In this study, we simulated the respiratory system with a computational model that carefully reflected its nonlinear characteristics. The model is an electrical circuit analogue, in which nonlinear resistors (R's), capacitors (C's), and inductors (L's) represent airway resistance, lung capacitance, and inertance of air, respectively. Just like in a living organ, the driving force of this model is P_{ pl }which is a superposition of HFCC pulses on the spontaneous breathing effort. We compared the simulation data to the in vivo data to demonstrate and understand the characteristics of airflow signal oscillation.
Methods
Conceptual model of the respiratory system
Our modelling and simulation for reproducing airflow signals required simplification of airway structure in the lung. The geometrical and dimensional structures of the airways were proposed by several researchers, among which Horsefield et al. [10] and Weibel [31] dissected and measured the human lungs and airways, and more recently, and Tawhai et al. [11] and Kitaoka et al. [14] proposed algorithmic approaches for reconstructing the branching structure. We employed Weibel's morphometry of the lung, which provides the geometries and the dimensions based on the symmetric dichotomous structure of airway branching when the lung volume is assumed to be 75% of the total lung capacity (TLC). According to his morphometry, the trachea is defined as airway generation 0 and it is separated into two geometrically identical daughter branches. Each daughter is repeatedly branched up to 22 times, thus the lung is considered to have 24 (0–23) airway generations and the number of airway branches total ~1.7 × 10^{7}. In each airway generation, there are 2^{ z }(z is a generation number) identical branches whose lengths and radii are provided and the dimensions of airway branches in each generation differ from generation to generation. He also suggested that airway generations 0–16 comprise the conducting zone, whereas airway generations 17 – 23 make up the respiratory zone in which gases are exchanged.
Our conceptual model as a summary of the simplification, 'the upper airway + 17 conducting airway generations + the alveolar space', is demonstrated in fig. 1 (a), and circuit analogue converted from fig. 1 (a) is shown in fig. 1 (c).
Dimensions of airway branches
Since the dimensions of every single airway branch in the lung vary during respiration, it is crucial to track them to extract proper values of R's, C's, and L's. The radii of the airway branches in airway generation Z are determined by transmural pressure (P_{ tmZ }), pressure difference between the pleural cavity and inside airway generation Z. In order to determine P_{ tmZ }, we need to rely on the study about transpulmonary pressure (P_{ tp }), pressure difference between the pleural cavity and alveolar space. According to Salizar et al. [26], P_{ tp }is a function of lung volume (LV) and TLC. With given LV, P_{ tp }is obtained by,
P_{ tp }= log (1  LV / TLC) × 7.22. (1)
If LV is functional residual capacity (FRC) and no airflow exists in the airways, the lung is at rest. In this situation, air pressure at any site is the same and atmospheric (zero), therefore P_{ tmZ }and P_{ tp }are the same. Our simulation begins with assuming that the lung was at rest. The equation of Lambert et al., or Lambert's tube law [16], is a function of P_{ tmZ }, and gives the ratio of the cross sectional area (A_{ Z }) to the maximum cross sectional area of an airway branch (A_{ maxZ }) in airway generation Z. That is,
A_{ Z }/ A_{max Z}= 1.0  (1.0  α_{0})(1.0  P_{ tmZ }/ P_{0})^{N} (2)
where α_{0}, α_{0}' and N are constants for each airway generation given from Lambert's tube law, and P_{0} = (α_{0}1)N/α_{0}'. As mentioned earlier, the dimensions of the airway branches provided by Weibel's morphometry are based on 75 % of TLC. Let A_{ Z75 }be Weibel's cross sectional areas in airway generation Z. By using
A_{ maxZ }for each airway generation can be found. Since A_{ maxZ }does not change whether or not the lung is at rest, the equation is valid to find out A_{ Z }during respiration.
The length of an airway branch is assumed to vary in away that;
l_{ z }/ l_{ FRC }= r_{ z }/ r_{ FRC } (3)
where l_{ z }and r_{ z }are the length and the radius of an airway branch in airway generation Z, respectively. l_{ FRC }and r_{ FRC }are the length and the radius on FRC, respectively.
Equivalent circuit elements modelling
The RCL Tnetwork consists of two resistors (R_{ gZ }), two inductors (L_{ gZ }), a capacitor (C_{ gZ }), and a DC voltage source (P_{ tm_ini }). R_{ gZ }and L_{ gZ }are the airway resistance and the inertance of air in the entire airway generation Z, respectively. C_{ gZ }, the airway capacitance of the entire airway generation Z, represents inflation and deflation of the nonrigid airway wall during respiration. P_{ tm_ini }is the initial value of P_{ tmZ }. This initial pressure counterbalances P_{ pl }, which is negative when the lung is at rest.
The values of the circuit elements in the RCL Tnetwork of airway generation Z can be obtained by totalling the 2^{ z }airway branches. Therefore, R_{ gZ }is given by
Similarly, L_{ g }z is given by
and C_{ gZ }is
C_{ gZ }= C_{ sZ }× N_{ z }[ml/cmH_{2}O], (6)
where R_{ sZ }, L_{ sZ }, and C_{ sZ }indicate the values of a single airway branch in airway generation Z. N_{ z }is 2^{ z }, the number of airway branches in generation Z.
R_{ sZ }depends on the classic Poiseuille equation and the Zeta correction factor proposed by Pedley et al. [21]. The Zeta correction factor is given by
and
R_{ e }is Reynolds number , V is air velocity, ρ is the density of air, and η is the viscosity of air.
Because electric current flows at the speed of light in an electrical circuit, it is necessary to compensate for the incomparably slower behaviour of airflow in the airways by placing inductors. Inertance of air in a single airway branch (L_{ sZ }) depends on the dimension of the airway branch [5] and it is given by
To compute the values for C_{ sZ }in equation (2), relation between A_{ Z }and P_{ tmZ }in airway generation Z, is used again. That is,
The alveolar space is represented by a capacitor in the circuit analogue. Alveolar capacitance (C_{ as }) is the ratio of the alveolar volume (ΔAV) to ΔP_{ tp }. Lung volume (LV) consists of airway volume and alveolar volume, and airway volume is negligible compared to alveolar volume. Therefore,
A resistor and an inductor characterize the upper airway, from the nasal/oral cavity to larynx. The equations for the resistance of the upper airway (R_{ ua }) that Jackson et al. [12] validated are described below:
where F_{ ua }is the airflow rate in the upper airway. As Marchal et al. [18] estimated, the value of the inductor that represents inertance of air is 0.00003 [cmH_{2}O·s^{2}/ml].
Numerical methods for nonlinear circuit analogue
where v_{ C }is the voltage difference between each end of the capacitor. Note that R(•) is a function of i and v_{ C }, and L(•) and C(•) are functions of v_{ C }just like in fig. 1 (c). Using backward Euler approximation [22], equation (14) is converted to a difference equation:
where Δt is time difference between sequence [n] and [n1]. Equation (15) is the same as
In equation (17), the right hand side consists of all known values, and the left hand side is a function of v_{ C }[n]. Suppose that v_{ C }[n] is x, equation (17) can be expressed as
f(x) = c, (18)
where c is a constant. Equation (18) can be easily solved using an iteration method [24]. To do this computation, MATLAB (Mathworks, Natick, MA) codes were written.
Protocols for measurement airflow signals
The subject sat upright on a chair for measuring and recording the airflow signals at the mouth with an inhouse built electronic spirometer. The subject worn a nose clip and breathed through a mouthpiece. After the HFCC device is properly set up, the compression pulses were applied and then the subject made several slow and large, but not to TLC, breaths. During the breaths the subject hold his glottis open until data collection was completed. This protocol was followed for the low (5 Hz), high (21 Hz), and medium frequencies (15 Hz) of HFCC pulses. Before each frequency recording of airflow signals, the subject rested for one minute. To ensure that the subject had adapted to HFCC pulses and had reached a steady state, only the last ten seconds of the one minute breathing were recorded and analyzed for our study.
Results
For the simulation, the parameters of the lung were determined to represent a normal healthy lung. The TLC, FRC, and RV of the model lung are 6000 ml, 2700 ml, and 1000 ml, respectively. The ambient atmospheric pressure is assumed to be zero, and the air density and viscosity are 0.00113 g/cm^{3} and 0.00019 g/cm·s, respectively. It was assumed that there would be no airflow in the airways and that initially the LV would be the same as the FRC. Change of P_{ pl }is the primary driving force of respiration and is initially 4.32 cmH_{2}O, which is the counterbalance to initial P_{ tp }given by equation (1).
From the measured values and the simulation of airflow signals at the mouth, we could observe two important phenomena during respiration pertaining to the airflow signal oscillation generated by HFCC. First, the amplitudes of HFCC airflow signals in phases I and III were smaller than those in phases II and IV. Second, the amplitudes in phases I and III became even smaller as the spontaneous airflow signal became greater.
Discussion
The number of the airway branches is estimated to be about 17 million [31], and the dimensions of each airway branch vary while the lung is being inflated and deflated during respiration. To reduce the computational burden, models of the respiratory system are often simplified by ignoring the nonlinear natures of the respiratory system [7, 8]. Linearization entails considerable flexibility in numerical analysis since linear circuits are computationally much cheaper than nonlinear circuits, either on a timedomain or a frequencydomain basis. However, even during the slowest breathing, airway resistance, lung capacitance, and inertance of air change due to the variation of airway dimensions as well as turbulence. In particular, increase of the airway resistance due to turbulence is so drastic that inaccurate evaluation of airway resistance may result in a misleading simulation result. Our simulation results demonstrated that such errors are more likely when HFCC intervention is applied. Evidence for this was that the result from a linear model was significantly different from that from a nonlinear model (fig. 8 and 9). As parts of the respiratory system, lung capacitance and inertance of air are also nonlinear although it was observed that their nonlinearity did not cause the nonuniform amplitudes of HFCC airflow signals. However, this does not imply that the nonlinear characteristics of lung capacitance and inertance of air have no role in the simulation of the respiratory system. Dimensional changes of the airway branches in the lung are responsible for the nonlinearity of lung capacitance and inertance of air. When dimensions of the airways are involved for simulation and prediction, the nonlinearity is very important for accurate results. For example, in the study of Sohn et al. [27], our model is used for estimation of air velocity, which is airflow rate ÷ cross sectional area. Since the cross sectional areas of airway branches do not vary linearly, lung capacitance and inertance of air also should not be linear in order to avoid discrepancy.
Ideally, studies about airflow in the airways would be best resolved by CFD (Computational Fluid Dynamics), however, using a CFD approach in this study presents several problems that cannot be overcome by modern technology. First, the turbulence mechanism is not completely known [23]. Even during the slowest breathing manoeuvre, turbulence exists in the proximal airways, and ignoring turbulence would not give true simulation results. Another obstacle is that the whole lung cannot be taken into consideration even with the latest supercomputing power. A CFD approach to simulate dynamics of airflow interacting in a huge number of airway branches in the lung requires extremely massive computation. There is no way to deal with it, if any, its computation time would be incredibly long. It should be also pointed out that CFD techniques for airflow in nonrigid wall tubes are not yet mature. Although not perfect, our circuit analogue model allows us to effectively simulate the nonlinear characteristics of the respiratory system.
It is well known that respiratory system impedance consists of airway impedance and chest wall (tissue) impedance [2–4]. Since the driving force of airflow in our model is the intrapleural pressure altered by HFCC pulses transferred from the body surface, our model does not necessarily incorporate chest wall (tissue) impedance. However, it is not clear how effectively the chest wall transfers HFCC pulses to the pleural cavity. Other factors should also be considered for impedance between the jacket of HFCC and the pleural cavity – such as clothes, posture, and tightness of the jacket. Currently, based on observations of Milla et al. [19], we assume that HFCC pulses on the body surface are transferred to the pleural cavity without any distortion of pulsation waveforms although some attenuation may exist.
Conclusion
In this study, the airflow signals measured at the mouth during use of HFCC are viewed as a composite of two causes: the spontaneous breathing effort and HFCC pulses. However, since the respiratory system is nonlinear, airflow signals at the mouth are not a mere superposition of the effects from these two causes. In laboratory measurements, the amplitudes of the airflow signal oscillation varied considerably despite the uniformity of the HFCC pulses. After confirming that the simulation results matched up with the observations, we analyzed the simulation data to explain the observed inconsistency in the HFCC airflow signal amplitudes. We found that the amplitudes of HFCC airflow signals behave as a function of spontaneous airflow signals. This is due to the nonlinearity of the respiratory system, particularly variations in airway resistance.
The findings in this paper may not be immediately applicable for HFCC therapy, but they do lead to ways to better prescribe HFCC therapy. Most importantly, the usefulness of our computational simulation and the modelbased approach as a tool to understand clinical observations of HFCC was well demonstrated in this paper.
List of abbreviations
 HFCC:

High Frequency Chest Compression
 FRC :

Functional Residual Capacity
 TLC :

Total Lung Capacity
 RV :

Residual Volume
 LV :

Lung volume
 AV :

Alveolar volume
 R:

resistor
 C:

capacitor
 L:

inductor
 A _{ Z } :

Airway cross sectional area in airway generation Z
 A _{ maxZ } :

Maximum Airway cross sectional area in airway generation Z
 A _{ Z75 } :

Airway cross sectional area in airway generation Z when LV is 75% of TLC
 C _{ as } :

Alveolar capacitance
 C _{ gz } :

Total airway capacitance in airway generation Z
 C _{ sZ } :

Single airway capacitance in airway generation Z
 F _{ inZ } :

Incoming airflow in airway generation Z
 F _{ outZ } :

Outgoing airflow in airway generation Z
 F _{ ua } :

Airflow in the upper airway
 L _{ gz } :

Total inductance (inertance of air) in airway generation Z
 L _{ sZ } :

Single inductance (inertance of air) in airway generation Z
 l _{ Z } :

The length of an airway branch in airway generation Z
 N _{ Z } :

Number of airway branches in airway generation Z
 P _{ alv } :

Alveolar pressure
 P _{ awZ } :

Airway pressure in airway generation Z
 P _{ pl } :

Pleural pressure or intrapleural pressure
 P _{ tm_ini } :

The initial value of transmural pressure
 P _{ tmZ } :

Transmural pressure in airway generation Z
 P _{ tp } :

Transpulmonary pressure
 P _{ tp_ini } :

The initial value of transmural pressure
 R _{ gz } :

Total airway resistance in airway generation Z
 R _{ sZ } :

Single airway resistance in airway generation Z
 r _{ z } :

The radius of an airway branch in airway generation Z
Declarations
Authors’ Affiliations
References
 Axe JR, Abbrecht PH: Analysis of the pressurevolume relationship of excised lungs. Ann Biomed Eng 1985, 13: 101–117.View ArticleGoogle Scholar
 Black LD, Dellaca R, Jung K, Atileh H, Israel E, Ingenito EP, Lutchen KR: Tracking variations in airway caliber by using total respiratory vs. airway resistance in healthy and asthmatic subjects. J Appl Physiol 2003, 95: 511–518.View ArticleGoogle Scholar
 Dellaca RL, Aliverti A, Lutchen KR, Pedotti A: Spatial distribution of human respiratory system transfer impedance. Ann Biomed Eng 2003, 31: 121–131. 10.1114/1.1541012View ArticleGoogle Scholar
 Dellaca RL, Black LD, Atileh H, Pedotti A, Lutchen KR: Effects of posture and bronchoconstriction on lowfrequency input and transfer impedances in humans. J Appl Physiol 2004, 97: 109–118. 10.1152/japplphysiol.00721.2003View ArticleGoogle Scholar
 Elad D, Shochat A, Shiner RJ: Computational model of oscillatory airflow in a bronchial bifurcation. Respir Physiol 1998, 112: 95–111. 10.1016/S00345687(98)00005XView ArticleGoogle Scholar
 Fink JB, Mahlmeister MJ: Highfrequency oscillation of the airway and chest wall. Respir Care 2002, 47: 797–807.Google Scholar
 Gillis HL, Lutchen KR: How heterogeneous bronchoconstriction affects ventilation distribution in human lungs: a morphometric model. Ann Biomed Eng 1999, 27: 14–22. 10.1114/1.161View ArticleGoogle Scholar
 Golden JF, Clark JW Jr, Stevens PM: Mathematical modeling of pulmonary airway dynamics. IEEE Trans Biomed Eng 1973, 20: 397–404.View ArticleGoogle Scholar
 Hansen LG, Warwick WJ, Hansen KL: Mucus transport mechanisms in relation to the effect of high frequency chest compression (HFCC) on mucus clearance. Pediatr Pulmonol 1994, 17: 113–118.View ArticleGoogle Scholar
 Horsfield K, Dart G, Olson DE, Filley GF, Cumming G: Models of the human bronchial tree. J Appl Physiol 1971, 31: 207–217.Google Scholar
 Howatson Tawhai M, Pullan AJ, Hunter PJ: Generation of an anatomically based threedimensional model of the conducting airways. Ann Biomed Eng 2000, 28: 793–802. 10.1114/1.1289457View ArticleGoogle Scholar
 Jackson AC, Milhorn HT Jr: Digital computer simulation of respiratory mechanics. Comput Biomed Res 1973, 6: 27–56. 10.1016/00104809(73)90061XView ArticleGoogle Scholar
 King M, Rubin BK: Physiological bases for the control of mucous hypersecretion. In Airway secretion. Edited by: Takishima T, Shimura S. NY: Marcel Dekker, Inc; 1994.Google Scholar
 Kitaoka H, Takaki R, Suki B: A threedimensional model of the human airway tree. J Appl Physiol 1999, 87: 2207–2217.Google Scholar
 Krumpe PE, Evrensel CA, Hassan AA: Superimposed oscillation enhance the clearance of mucus stimulant at low air flows in a rigid tracheal model. ASME International Mechanical Engineering Congress & Exposition 2002, 437–438.Google Scholar
 Lambert RK, Wilson TA, Hyatt RE, Rodarte JR: A computational model for expiratory flow. J Appl Physiol 1982, 52: 44–56.Google Scholar
 Lapin CD: Airway physiology, autogenic drainage, and active cycle of breathing. Respir Care 2002, 47: 778–785.Google Scholar
 Marchal F, Haouzi P, Peslin R, Duvivier C, Gallina C: Mechanical properties of the upper airway wall in children and their influence on respiratory impedance measurements. Pediatr Pulmonol 1992, 13: 28–33.View ArticleGoogle Scholar
 Milla CE, Hansen LG, Weber A, Warwick WJ: Highfrequency chest compression: effect of the third generation compression waveform. Biomed Instrum Technol 2004, 38: 322–328.Google Scholar
 Nucci G, Suki B, Lutchen K: Modeling airflowrelated shear stress during heterogeneous constriction and mechanical ventilation. J Appl Physiol 2003, 95: 348–356.View ArticleGoogle Scholar
 Pedley TJ, Schroter RC, Sudlow MF: Flow and pressure drop in systems of repeatedly branching tubes. J Fluid Mech 1971, 46: 365–383.View ArticleGoogle Scholar
 Pillage TL, Rohrer RA, Visweswariah C: Electronic circuit and system simulation methods. McGrowHill; 1995.Google Scholar
 Rajagopal KR: On some unresolved issues in nonlinear fluid dynamics. Russ Math Surv 2003, 58: 319–330. 10.1070/RM2003v058n02ABEH000612View ArticleGoogle Scholar
 Robinson RC: An introduction to dynamical systems: Continuous and discrete. NJ: Pearson Prentice Hall; 2004.Google Scholar
 Rodarte JR, Rehder K: Dynamics of respiration. In Handbook of Physiology. 2nd edition. Bethesda, MD: American Physiological Society; 1986.Google Scholar
 Salazar E, Knowles JH: An Analysis of PressureVolume Characteristics of the Lungs. J Appl Physiol 1964, 19: 97–104.Google Scholar
 Sohn K, Holte JE, Phillips JR, Warwick WJ: Modeled velocity of airflow in the airways during various respiratory patterns. the 26th Annual International Conference of the IEEEEMBS; San Francisco, CA 2004, 3925–3928.Google Scholar
 Tomkiewicz RP, Biviji A, King M: Effects of oscillating air flow on the rheological properties and clearability of mucous gel simulants. Biorheology 1994, 31: 511–520.Google Scholar
 Warwick WJ, Hansen LG: Chest compression apparatus. U.S. Patent 4 838 263 June 13, 1989
 Warwick WJ: Mechanisms of mucous transport. Eur J Respir Dis Suppl 1983, 127: 162–167.Google Scholar
 Weibel ER: Morphometry of the human lung. Berlin: Springer; 1963.View ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.