You are on page 1of 16

CHAPTER 54

MEASUREMENT TECHNIQUES IN RESPIRATORY MECHANICS


Jason H. T. Bates

he assessment of respiratory mechanics begins with the measurement of variables whose correlations contain information about the mechanical functioning of the respiratory system. These variables are usually pressures and flows, or volumes, of gas made at appropriate sites. The amount of detail that can be gleaned about the mechanical function of the respiratory system depends on which particular variables are measured and the conditions under which they are measured. Although the modern history of measurement in respiratory mechanics extends back at least 100 years, novel methods and approaches are still being developed as advances in instrumentation and computer technology continue to extend the boundaries of what is possible. This review of measurement in respiratory mechanics begins with the general theory of measurement as it applies to modern electronic transducers and the acquisition of data by digital computer. It then proceeds to consider how this theory is applied to the measurement of respiratory pressures, flows, and volumes. Finally, it is shown how these measurements are used in combination to collect information from which respiratory mechanics are determined.

this is never the case in practice. It is therefore crucial to understand the imperfections of the transducer to be used in any particular application and whether or not these imperfections are going to limit the amount of information that can be extracted from the measured signals. Static Properties of Transducers We first consider the various static properties that characterize a transducers performance. These are properties that do not depend on how rapidly the measured signal is varying, which is equivalent to saying that the transducer has no trouble keeping up with the signal. One such property is linearity, which refers to the extent to which the voltage v produced by a transducer can be represented in terms of the biologic signal s by an equation of the form v as b (54-1)

MEASUREMENT THEORY
The general measurement situation is depicted in Figure 54-1, which shows the steps involved in converting a biologic signal into a string of numbers stored on a computer. At each step we have the potential for errors to occur, represented as additive noise. First we consider the issues involved in transducing the biologic signal into an electrical signal. This is performed by a transducer. We then consider the process of digitizing the electrical signal so that it can be manipulated in digital form on a computer, which is how any subsequent data analysis is carried out.

TRANSDUCERS
A transducer is something that converts energy from one form to another, although for the present purposes we restrict ourselves to the more particular definition of the conversion being between some biologic quantity we are interested in and a voltage. Ideally, we would like the voltage to be a perfect representation of the biologic quantity, but

where a and b are constants (solid line in Figure 54-2). Manufacturers of transducers usually specify the linearity of a transducer in terms of its full-scale output. A value of 1% is typical. Although linearity is a desirable trait, with the availability of digital computers it is not essential. Suppose, for example, that v is not a linear function of s as in Equation 54-1 but is instead a curvilinear function of s that can be accurately represented as a polynomial (dashed line in Figure 54-2). If the polynomial coefficients are known (say, from a prior calibration experiment) it is a simple matter to use a computer to invert the equation to obtain s as a function of v, allowing the recorded voltage to be related back to the value of the original biologic signal. Another important static property of a transducer is hysteresis, which refers to the extent to which the voltage corresponding to a biologic signal differs depending on whether the immediately preceding value was below or above the current value (Figure 54-3). Hysteresis is obviously a bad thing and is unfortunately extremely difficult to correct for, so one should always try to select a transducer with minimal hysteresis. In selecting a transducer for a particular application, it is important to make sure it has the appropriate resolution and dynamic range, which determine the smallest and largest changes in the biologic signal that can be accurately

624

Clinical Respiratory Physiology


Noise Noise

Biologic signal

Voltage Transducer

Analog-digital converter

Digital signal

Computer

Voltage

FIGURE 54-1 The general measurement scenario.

measured. Related to the issue of resolution is the signal-tonoise ratio. What the transducer in Figure 54-1 actually records in response to a biologic signal is a voltage plus noise. The latter must be substantially lower than any important changes that are to be recorded in the biologic signal (by preferably at least an order of magnitude). Dynamic Properties of Transducers We now consider the somewhat more complicated issue of dynamic properties of transducers. These refer to those characteristics that determine how well a transducer can respond to a changing biologic signal. Most transducers are low-pass systems because they can respond faithfully to slowly varying signals but have increasing difficulty keeping up as the signal increases in frequency. Some transducers are high-pass systems because they faithfully record high frequencies but do not respond when frequency is very low. Transducers are characterized in general by the way in which they respond to input signals that vary sinusoidally. Provided that the transducers are linear, their voltage outputs to an input sine wave will also be sinusoidal with the same frequency but will in general be altered in amplitude by a factor A and shifted in phase by an amount . That is, if the input sine wave is sin(2ft), then the output will be Asin(2 ft ) (Figure 54-4). The values of A and depend on frequency and together constitute the frequency response of the transducer.

Biologic signal

FIGURE 54-3 Hysteresis. The voltage output by a transducer in response to a particular value of the input biologic signal depends on whether the value was approached from above or below.

DIGITAL DATA ACQUISITION


Once a transducer has converted a biologic signal into a voltage, it must be recorded in some permanent medium for subsequent analysis and display. In the early days of physiology (ie, until about 20 years ago), physiologic recording was achieved by have a writing instrument move laterally

over a suitable recording medium as it scrolled by. In the early part of the twentieth century, the writing instrument was a rigid pointer and the scrolling medium was a cylindrical drum covered in soot. This was eventually replaced by the electronic chart recorder, consisting of one or more ink pens writing on a roll of paper moving past at constant speed. Each experiment produced a (frequently large) stack of paper that then had to be analyzed manually if calculations were to be made from the recorded signals. Since the advent of the modern laboratory digital computer, however, these earlier analog recording devices have been replaced by digital computers that both record and analyze experimental data. The speed and flexibility of computers, together with their universal availability, have revolutionized the way that physiologic research is done. Everything is thus now done digitally, beginning with the recording of the analog voltage signal arriving from the transducer. Conversion of an analog signal to digital form is known as digitization and is accomplished with an analog-to-digital (A-D) converter. Figure 54-5 shows what is involved. The analog (continuous) voltage signal is sampled at regularly spaced intervals by the A-D converter, and the resulting set of voltage values is stored in the computer as a set of numbers. These numbers and their locations in time then constitute our representation of the original signal. Resolution and Discretization Error A major consideration when digitizing signals is the resolution of the A-D converter, which defines the smallest difference in voltage level that it can distinguish in the incoming analog signal. A-D

Voltage

Input

Output

1.0

Transducer

Biological signal

FIGURE 54-2 Characteristics of a linear (solid line) and nonlinear (dashed line) transducer.

FIGURE 54-4 If an input to a linear transducer is sin(2ft), its output in general will be Asin(2ft ), where the values of A and characterize the frequency response of the transducer.

Measurement Techniques in Respiratory Mechanics


0.21 0.43 0.50 0.50 0.42 0.33 0.33 0.37 0.48 0.52 0.59 0.58 . . .

625

V(t)

0 t

FIGURE 54-5 Digitization. An analog voltage signal V(t) is sampled every t seconds to produce a series of numbers that are stored in a computer for subsequent analysis.

converters are set up to receive voltages within a specified range, such as 10 to 10 volts. This voltage range is divided into N equally spaced bins numbered 0 to N 1 (Figure 54-6). N is determined by the number of bits in the A-D converter. A 12-bit A-D converter has 212 4,096 bins, a 16-bit converter has 216 65,536 bins, and so on. Thus, for example, a 12-bit converter with a 10-volt input range can resolve voltage differences of 20/4,096 0.0049 volts. The finite resolution of an A-D converter means that care must be taken to ensure that it is able to resolve the smallest differences in voltage required by the experimenter. For example, suppose one is measuring flow of gas entering a patients lungs during mechanical ventilation. If the flow reaches 2,000 mL.s1 during both inspiration and expiration, then the range of flows encountered is 4,000 mL.s1. If this flow signal is recorded on a 12-bit A-D converter, the resulting digitized signal has a maximum resolution of 4,000/4,096 0.98 mL.s1. However, it only achieves this resolution if the entire dynamic range of the A-D converter is used. This only occurs if the analog signal produced by the flow transducer is amplified so that a flow of 1 2,000 mL.s produces the lowest voltage the A-D converter can receive (eg, 10 volts), and similarly a flow

of 2,000 mL.s1 produces the highest receivable voltage (eg, 10 volts). This may not be the case. A common error in the laboratory occurs when the voltage signal coming in from the transducer is not amplified enough to make use of many of the discrete bins in the A-D converter. As an example, suppose that the flow signal of 2,000 to 2,000 mL.s1 results in a voltage signal that only occupies the range of 0.1 to 0.1 volts. Now only the middle 1% of the available bins (numbers 2,028 to 2,069) of the A-D converter are used, giving a 100-fold reduction in flow resolution. In extreme examples of this situation, the discrete levels of the A-D converter will be apparent in a plot of the resulting voltage signal (Figure 54-7). The errors incurred in having insufficient vertical resolution in an A-D converter are called discretization errors. Sampling Theorem and Aliasing Another major question that arises when digitizing analog signals is how frequently to sample the signal. A continuous signal is composed of an infinite number of infinitesimally spaced points yet must somehow be represented by a finite number of digitized values. Obviously, if a signal is changing rapidly, then sampling its value infrequently will cause the detail between the samples to be lost. On the other hand, sampling a long signal too rapidly might result in an unmanageably large number of data points. Thus, one might be inclined to think that the choice of sampling rate represents a tradeoff between capturing detail in the original signal on the one hand and avoiding being overwhelmed by the volume of data on the other. Fortunately, however, we are saved by something known as the sampling theorem (often associated with engineers Shannon and Nyquist). To understand the sampling theorem, it is necessary to first understand what is meant by the frequency content of a signal. Any analog signal can be expressed as the sum of a series of sine wave functions of appropriate frequency, amplitude, and phase. Furthermore, this collection of frequencies, amplitudes, and phases is unique to that signal, which means they define it distinctly from all other possible signals. A signal can be decomposed into its individual sine wave components by the Fourier transform. A simple example is given in Figure 54-8, which shows the four sine waves making up a signal that looks considerably more complex than any of its individual components. The frequency content of a signal thus refers to the unique spectrum of

Analog signal +10 volts

Digitized signal 4095

-10 volts

FIGURE 54-6 Digitization of the 10 volt analog range by a 12-bit A-D converter. The 20-volt span of possible input signals is assigned to 4,096 numbers representing successive increments of 4.9 millivolts.

FIGURE 54-7 Discretization error. The left panel shows an analog signal faithfully captured by a set of digitized points. The right panel shows points occupying discrete levels corresponding to adjacent bins of the A-D converter.

626

Clinical Respiratory Physiology

Fourier transform

the intervening continuous signal segments by fitting a sine wave to the points. Thus, by sampling an analog signal at or above the Nyquist rate, we lose no informationthe entire original continuous signal can be reconstructed from the sampled points alone. The sampling theorem in principle means we lose nothing in moving from a continuous to a digital environment. However, there is a practical issuewhat if f0 is so high that we still end up with an unmanageably large number of data points when we satisfy the sampling theorem? Unfortunately, we cannot drop the sampling rate below 2f0 and hope to merely sacrifice some detail in the retained samples. The high frequencies in the analog signal do not simply disappearthey turn up in the digitized data but at wrong (much lower) frequencies! This is a particularly insidious problem known as aliasing, and experimenters must always be careful to avoid it, especially if the power spectra of the collected data are of interest. Aliasing can lead to completely erroneous spectral characterization of a signal and at the very least degrades the signal-to-noise ratio of a measurement. Figure 54-9 shows how aliasing occurs; a sine wave is sampled at a frequency below the Nyquist rate to yield a set of points that look as if they represent a different sine wave of a much lower frequency. Obviously, aliasing can be avoided by sampling at or above the Nyquist rate. However, in practice it is usually necessary to force the analog signal to be band limited to some manageable f0. This is done by passing the signal through a high-quality low-pass electronic filter before it is digitized. Such filters are known as antialiasing filters. It is important to remember that filtering for antialiasing must be done prior to digitization of the signal. Once aliasing occurs, no amount of digital filtering after the fact can fix the problem.

FIGURE 54-8 An analog signal can be decomposed (by the Fourier transform) into its component sine waves, each with its own frequency, amplitude, and phase.

sinusoidal components from which it is composed. In particular, if the frequency content of a signal is such that none of its sinusoidal components has a frequency greater than some maximum frequency, then the signal is said to be band limited. The sampling theorem says that it is possible to capture all the information in a band-limited analog signal by sampling the signal at a rate at least twice the highest frequency in the signal itself. Thus, for example, if Fourier analysis of a signal reveals that it contains no components with frequencies greater than f0, then we need not sample the signal any faster than 2f0 (known as the Nyquist rate). Intuitively, this makes sense because sampling a sine wave at its Nyquist frequency means collecting another sample from the signal every time it changes direction. If we know we are sampling a sine wave, then from this set of points we can reconstruct

FIGURE 54-9 A sine wave (solid line) is sampled (dots) below the Nyquist rate, yielding a set of data points defining a different sine wave (dashed line) at a lower frequency.

Measurement Techniques in Respiratory Mechanics

627

MEASUREMENT OF RESPIRATORY SIGNALS


Having established the general considerations for measuring signals in the laboratory, we now move on to consider those signals specific to the study of mechanical lung function.

PRESSURE
The measurement of pressure is central to the study of respiratory physiology because it is pressure that generates the flow of gas needed to ventilate the lungs. Pressure Transducers Pressure transduction is based on the graded deformation of some mechanical element whose altered configuration is read by some electronic means (Figure 54-10). Until about 15 years ago, the mainstay of pressure measurement in the respiratory physiology laboratory was the variable reluctance transducer in which a thin metal disk is placed between the primary and secondary coils of a transformer excited by several kHz of alternating electric current. A pressure difference either side of the disk causes it to deform in a way that alters the magnetic flux linkage between the transformer coils, thereby changing the induced voltage in the secondary coil. The change in voltage is then transformed into a DC voltage proportional to the pressure difference. These transducers are sensitive and accurate. They also typically have a frequency response that is flat to 20 Hz or more, depending on the length of the tubing connecting its ports to the sites of pressure measurement.1 However, they are somewhat cumbersome and can be damaged by over pressurization. In recent years, respiratory pressure measurement has been taken over by the piezoresistive transducer,2 in which the pressure-sensitive element changes its electrical resistance as it deforms. If a constant voltage (or current) is passed through the piezoresistive element when it is configured to be one of the four arms of a suitably balanced Wheatstone bridge, the voltage across the bridge is then proportional to the change in the elements resistance. A medium-gain amplifier followed by an antialiasing filter are the only remaining elements required to produce an electrical signal proportional to pressure that is ready for digitization. When piezoresistive pressure transducers were first used in respiratory physiology in the 1980s, they tended to suffer from baseline drift, were affected by orientation and temperature, and were not very sensitive. These problems have now been essentially overcome allowing piezoresistive transducers to be exploited for

their several advantages. These include an extremely high-frequency response (typically flat to several hundred Hz), robustness (they can be pressurized to many times their nominal full-scale range without damage), and the fact that they can be manufactured using solid-state technology to be very small and light. Piezoresistive transducers are also much cheaper than their variable reluctance counterparts and require simpler electronic signal conditioning circuitry. Measuring Pressure at the Airway Opening The assessment of pulmonary function frequently requires that the pressure in a flowing stream of gas be measured, such as at the entrance to the endotracheal tube in a mechanically ventilated patient. Gas always flows down a pressure gradient, so at each point along the stream of flowing gas there is a driving pressure pushing the gas downstream of it. The goal is to determine this driving pressure. The easiest way is to insert a perpendicular tap into the tube and connect it to a pressure transducer (Figure 54-11). This provides what is known as lateral pressure (Plat), and it corresponds to the pressure exerted perpendicular to the direction of flow as the gas moves past the point of measurement. It turns out, however, that Plat is less than the pressure driving the gas along the tube because of a phenomenon known as the Bernoulli effect, which occurs because of the principle of conservation of energy; the faster gas is moving along the tube, and consequently the larger its kinetic energy, the more it loses in potential energy, manifest as a drop in Plat. Plat underestimates true driving pressure, in a tube of cross-sectional area , by an amount Pb given by the formula A with flow V . V2 (54-2) Pb 2A2 where is the density of the gas and is a factor determined by the flow velocity profile. For example, if the profile is flat (the linear velocity of the gas molecules is the same at every point in the tube cross section) then 1 and 2 if the profile is parabolic. If the gas in the tube were stationary, then Plat would equal driving pressure, a condition that can always be achieved by connecting a pressure transducer to a small tube

Plat

Deformable element Pressure source

. V

FIGURE 54-10 Pressure transducer. The application of a pressure deforms an element whose configuration is converted into an electrical signal proportional to the degree of deformation.

Pstat
FIGURE 54-11 Lateral pressure (Plat) measured through a lateral tap, and static pressure (Pstat) measured with a Pitot tube.

628

Clinical Respiratory Physiology

that enters the flow stream laterally and then bends until its open end faces directly into the oncoming flow (see Figure 54-11). Such a device, called a Pitot tube, measures the static pressure (Pstat) in a small parcel of gas that has been brought to rest by abutting up against the tube opening. Pb is the difference between Pstat and Plat. That is, Plat Pstat Pb (54-3) The problem for respiratory pressure measurement caused by the Bernoulli effect is apparent from Equation 54-2, which shows that Pb depends on the square of flow divided by cross-sectional area. When the area is large enough, Pb is negligible. However, as the area decreases there comes a point at which Pb starts to become important, compared with Pstat. Indeed, Equation 54-3 shows that for small tube areas Plat may even become negative. Thus, it is important in any application in which lateral pressures are measured to be sure that the Bernoulli effect is not significantly affecting the measurement of the desired quantity, namely driving pressure.3 The Bernoulli effect may also be an important factor influencing the measurement of pressures at the distal end of an endotracheal tube in an intubated patient.4 Esophageal and Gastric Pressures Two other pressures of great practical importance in the study of respiratory physiology are those in the esophagus and stomach. Esophageal pressure (Pes) is a useful surrogate for pleural pressure, the esophageal lumen being separated from the pleural space by only soft tissue. Gastric pressure (Pga) measures the pressure exerted on the diaphragm by the abdominal contents, which is important for understanding active expiration. The difference between Pes and Pga is the pressure across the diaphragm, which is important in studies of respiratory muscle function.5 Both Pes and Pga can be measured using a balloon-tipped catheter (Figure 54-12). A thin-walled latex balloon, a few centimeters in length, is sealed over a thin plastic catheter, typically about 100 cm long. The balloon is passed into the esophagus, usually via the nose. Once in place in either the esophagus or stomach, a small volume of air is injected into the balloon and a pressure transducer is connected to the proximal end of the catheter. The volume of air in the balloon must be sufficient to prevent the walls of the balloon from occluding all the multiple holes in the end of the

catheter, but not so much that there is tension in the balloon walls. The correct placement of the balloon is gauged from the nature of the recorded pressure signals. When the balloon is in the stomach, spontaneous inspiration produces a positive deflection in the recorded Pga. As the balloon is withdrawn and enters the esophagus, the inspiratory pressure swings will suddenly become negative. At this point, when inspiratory efforts are made against an occluded airway (the so-called occlusion test), the deflections in Pes should match those in pressure measured at the airway opening (Pao). Thus, a regression of Pes versus Pao should yield a slope of unity.6 In practice, slopes that differ from 1.0 by up to 10% are common. Although the occlusion test requires that the subject be able to breathe spontaneously, it has been shown that the esophageal balloon also works well during paralysis.7 The frequency response of the esophageal balloon is obviously somewhat compromised by the fact that pressure changes in the esophageal lumen must be transmitted through the air inside a long thin catheter to a pressure transducer some distance away. However, a reasonably good response to 30 Hz has been observed.8 Pes has also been measured using catheter-tip piezoresistive transducers, which have been shown to perform well9 and have a much better frequency response than balloon catheter systems. In small animals, Pes can be measured using a water-filled catheter.10 Alveolar Pressure In experimental animals it is possible to measure alveolar pressure directly using a technique known as the alveolar capsule.11 The chest is opened and retracted to expose the pleural surface of the lung to which a small plastic capsule is fixed (Figure 54-13). The capsule has a cylindrical chamber leading down to a small window on the pleural surface encircled by a flange. If the pleural surface is first swabbed with alcohol and dried, the flange can be secured to it with cyanoacrylate glue. Small puncture holes are then made in the pleural surface within the capsule window. If the holes are made carefully to a depth of 1 to

Pressure transducer Capsule Pleural surface

10 cm 100 cm

Sub-pleural alveoli

Pressure transducer

Plastic catheter

Latex balloon

Terminal airways

FIGURE 54-12 An esophageal balloon catheter. Note the multiple holes in the walls of the catheter inside the balloon.

FIGURE 54-13 The alveolar capsule.

Measurement Techniques in Respiratory Mechanics


Differential pressure transducer

629

. V

Resistive element

FIGURE 54-14 The pneumotachograph.

response of a pneumotachograph degrades rapidly as the tubing connecting the transducer to the lateral taps either side of the resistance element increases in length. Inadequate frequency response can be corrected for to a significant extent by digital compensation, giving an effectively flat frequency response to more than 100 Hz.16 Another consideration for pneumotachographs concerns their dynamic common-mode rejection characteristics. If the two ports of the pneumotachograph are both subjected to the same change in pressure, ideally the device should register a flow of zero. This is not always the case, however, particularly if the physical dimensions of the two ports are different and if the input impedance of the pressure transducer is not very large compared with that of the system under investigation when flow oscillates at a high frequency. Digital compensation methods can improve the situation to a certain extent.17 Poor dynamic common-mode rejection is a significant problem for pneumotachographs used with very small animals.18 Other Devices for Measuring Flow Although the resistive pneumotachograph is the mainstay for measuring flow in respiratory applications, other devices have been used. For example, ultrasonic transducers based on differences in time-of-flight of sound propagating into the direction flow versus away from it have an excellent frequency response and avoid the problems of a resistive element becoming clogged with secretions.19 Devices based on the rate of cooling of a heated wire are also in use.20

2 mm, preferably with a cautery needle, bleeding is minimal and the subpleural alveoli are brought into contact with the capsule chamber. A small piezoresistive pressure transducer can then be lodged in the chamber to give a direct recording of subpleural pressure. In large animals, such as dogs, several alveolar capsules can be installed at different sites over the lung surface.1214 A single alveolar capsule can even be used in an animal as small as a mouse.15

FLOW
Pneumotachographs The mainstay of flow measurement in respiratory physiology is the pneumotachograph, which is a calibrated resistance (R) across which a differential pressure is measured (Figure 54-14). When gas flows through the pneumotachograph, there is a pressure drop (P) from the upstream side of the resistance to the downstream side ) increases, thus that increases as flow (V P RV (54-4) over the range of flows of interest If R is independent of V then the pneumotachograph is said to be linear. Manufacturers strive for linearity and always quote the linear range of a pneumotachograph. However, if the device is nonlinear, , it is a simple matter to invert so that R depends on V from a measureEquation 54-4 on a computer and calculate V is known. ment of P, provided that the relationship of R to V The frequency response of a pneumotachograph depends on the construction of its resistive element. Some consist of a honeycomb arrangement of conduits, whereas others consist of a wire screen. The honeycomb type is less likely to become partially blocked by secretions but has a poorer frequency response than the screen type. Either type should be heated to above body temperature during prolonged use to avoid breath condensate from settling on the resistive element and changing its resistance (and hence altering the calibration of the device). Pneumotachographs can have a good frequency response above 20 Hz with a resonance occurring at around 70 Hz, provided that the associated differential transducer has a response at least that good and is connected with the shortest possible lengths of tubing.1 The frequency

VOLUME
Direct Measurement of Volume The volume of gas entering the lungs can be measured directly with a spirometer attached to the mouth or from the pressure or flows emanating from a whole body plethysmograph when the subject breathes through a conduit connected to outside the plethysmograph. A more convenient but less accurate plethysmographic method is provided by the changes in trunk volume assessed with an inductance plethysmograph.21 Recently, a more accurate but expensive optical plethymograph has been developed that allows the detailed measurement of thoracic movement during breathing.22 However, for many applications requiring the assessment of lung function, the easiest way to assess changes in lung volume is to integrate the flow measured at the mouth with a pneumotachograph. Integration of Flow Before the advent of the modern laboratory digital computer, integration was typically achieved in real time using an electronic circuit based on the charging of a capacitor. Nowadays, integration is performed digitally on a computer. The digitized flow signal consists of a series 1,V 2,V 3, } separated by equal time intervals of data points {V t. A simple formula for numerical integration is the trapezoidal rule, which involves approximating the area under a section of a curve by a trapezoid. This means 2 by a straight line 1 and V approximating the curve between V (Figure 54-15) and calculating the area A under it as A V1 + V2 t 2 (54-5)

630

Clinical Respiratory Physiology

. V (t)
A1 A2

t1

t2

t3

FIGURE 54-15 Trapezoidal integration of flow involves connecting each sampled flow value by straight line and summing the areas under each trapezoid.

The area A under a section of curve is then simply the sum of all the individual As, thus A

Ai
i =1

V 1 + V2 + V3 + 2

+ Vn 1 +

Vn t 2

(54-6)

There are other more accurate, and more complicated, numerical integration schemes. However, the key thing is that t should be small enough so that the errors involved in approximating the true curve between points is negligible. This can be tested for by integrating the data using progressively smaller values for t, until the results do not change.

MEASUREMENT OF LUNG FUNCTION


Now that a set of respiratory data has been collected and stored on a computer, as described above, we are faced with the task of having to interpret it. How this is done depends very much on the nature of the data collected. In most cases, however, the assessment of lung function from respiratory data is based on some model idealization of the real system under study. The model is only useful if it behaves like the real system to a satisfactory degree. It is therefore crucial to understand what model is being invoked whenever lung function is being assessed so that the suitability of the model can be evaluated. In most applications, the model concerned is very simple, typically involving only a single compartment. In some cases, the model may be very complicated with multiple alveolar compartments and airway branches represented. Regardless of the level of model complexity, however, it is crucial that any investigation of mechanical lung function begin with an understanding of the mathematical model being invoked.

forced expiration, yields quantities that are usually treated as purely empirical. The forced vital capacity (FVC) and the forced expiratory volume in 1 second (FEV1) are widely used either individually or as a ratio to diagnose a variety of common lung pathologies. The great advantage of the forced expiration is that it can be performed easily in an outpatient setting, requiring nothing more than a device for measuring flow (or volume) and some cooperation from the subject. The measuring device must be able to measure flows of greater than 10 L.s1 with a frequency response that captures the rapid onset of flow at the start of a maximal forced expiration beginning from total lung capacity. There are commercially available systems designed specifically for this task, either by direct spirometric measurement of exhaled volume or by pneumotachograph measurement of flow. The clinical utility of forced expired flows arises from the phenomenon of flow limitation, whereby exhaled flow reaches a maximum value independent of further increases in expiratory muscle activity. Subjects with normal muscle function can reach this limiting flow over most of the vital capacity range. The limiting flow is reduced in obstructive lung disease, and the relationship of flow to exhaled volume assumes shapes that are characteristic of various lung pathologies. Detailed accounts of the flow-volume curve and its relation to pulmonary disease can be found in standard texts.23 Unfortunately, the underlying mechanical mechanisms in the lung responsible for producing the flowvolume loop are complex and nonlinear, so parameters such as FVC and FEV1 are not easily related to a model of the respiratory system. Consequently, making the link between function and lung structure from forced expired flows has not been straightforward. In other words, although parameters such FVC and FEV1 are sensitive to both obstructive and restrictive lung disease, it is unclear how to deduce the site or nature of an abnormality from the values of these parameters, although some sophisticated computer modeling studies have attempted to elucidate the relationships between the forced expired flow-volume loop and the physical properties of the lungs.24 Nevertheless, the convenience of measuring forced expired flows and their diagnostic utility make this technique the most clinically important of all lung function measurements.

METHODS BASED MODEL

ON THE

SINGLE-COMPARTMENT

The model of respiratory or pulmonary mechanics most frequently employed as the basis for lung function measurements is that of a single elastic compartment served by a single flow-resistive airway (Figure 54-16). This model assumes that the lungs are homogeneously ventilated and that all alveolar pressures are equal to each other at all times. Plethysmography Whole-body plethysmography complements the measurement of forced expired flows as the other major methodology for measuring lung function in current clinical use. Plethysmography is used to measure two important parameters of lung function, thoracic gas volume (TGV) and airway resistance (Raw), both under the assumption of the single-compartment linear model. The

FORCED EXPIRED FLOW


Having just expounded on the importance of models in the analysis of lung function data, it turns out that the traditional mainstay of clinical pulmonary function testing, the

Measurement Techniques in Respiratory Mechanics

631

FIGURE 54-16 The single-compartment model of the respiratory system. R is resistance and E is elastance (see Equation 54-8).

calculation of TGV and Raw using a body plethysmograph is covered in numerous other articles and books23 and so is described only briefly here. TGV is measured first by having the subject make breathing efforts against a closed airway while the change in airway opening pressure (Pao) is measured (Figure 54-17A). At the same time, the pressure in the plethysmograph or the flow leaving it (depending on whether it is a pressure or flow box) provides a measure of changes in total body volume (V). Assuming that the only compressive part of the body is the gas in the lungs, V is equal to the compression of the thoracic gas. Applying Boyles law gives Pao V 1000 TGV (54-7)

) is measured at the subjects the plethysmograph. Flow (V mouth with a pneumotachograph while pressure (Pbox) inside the plethysmograph is measured (Figure 54-17B). Pbox varies because of two factors. One of these is that gas inspired from the plethysmograph becomes humidified and heated to body temperature, thereby increasing in volume. This increases the combined volume of the remaining plethysmographic gas and the subjects body and so compresses the plethysmographic gas. The other factor is that a negative pressure in the thorax is required for inspiration and the converse for expiration. This results in compressive volume changes in the thoracic gas that are again manifest as changes in the combined volume of plethysmographic gas and subject. To the extent that the first source of changes in Pbox can be ignored, the changes in Pbox can be equated to compressive changes in thoracic gas volume, which can then be used, again via Boyles law, to infer the changes in alveolar pressure (Palv). The difference between Palv and Pao then yields a measure of Raw. divided by V The plethysmographic measurement of TGV and Raw assumes that the lungs are a homogeneously ventilated (ie, effectively single-compartment) system. This works well in normal lungs but may not in severely diseased lungs in which obstructed airways can lead to marked differences between Palv and Pao when panting against a closed airway.25 An unrestrained version of plethysmography has been used in an attempt to assess lung function in infants26 and small animals.27 This technique involves merely measuring variations in Pao as the subject breathes while inside the box. Although these pressure swings arguably give information about the pattern of breathing,28 they cannot provide meaningful information about respiratory mechanical function,29 despite recent claims to the contrary.30,31 Resistance and Elastance The single-compartment model of respiratory mechanics (see Figure 54-16) is described by a simple and extremely useful mathematical equation if it is linear. If the resistive pressure drop between one end of the airway and the other is considered to be pro ) through it, then the constant portional to the flow of gas (V of proportionality (R) is termed resistance. Similarly, the elastic recoil pressure inside the compartment is taken as

where we have equated 1 atmosphere to 1,000 cm H2O in the denominator on the left side of Equation 54-7, which can be rearranged to give a measure of TGV. Next, this measurement of TGV is used to estimate Raw by having the subject pant while completely enclosed inside

B . V Pao

Pao

Pbox

Pbox

FIGURE 54-17 Body plethysmograph used to measure A, TGV and B, Raw. Pao is airway opening pressure, Pbox is pressure inside the is mouth flow. plethysmograph, and V

632

Clinical Respiratory Physiology


2 Flow (l/s) Volume (l) Pressure (cmH2O)

proportional to the volume of the compartment above some elastic equilibrium volume, with its constant of proportionality (E) termed elastance. Finally, we must account for the possibility that the pressure (P) applied across the model (from the entrance to the airway through to the outside of the elastic compartment) has some finite value P0 when V and V are both zero. Simple addition of pressures shows that P is the sum of the resistive and elastic pressures, thus, (t) EV(t) P0 P(t) RV (54-8) , and V explicitly as where we have written the variables P, V functions of time (t) to remind us that they all vary during breathing. Equation 54-8 is used in respiratory investigations to estimate R, E, and P0 via multiple linear regression. This pro, and V that, when used in the right vides those values of P, V side of Equation 54-8, provide the closest approximation to the measured P in the left side. Closest here means in the least squares sense, which means that the sum of the squared differences between the measured value of P and the model values is minimal. R and E are taken as measures of the resistance and elastance of the lung or respiratory system, which in a sense they are. Strictly speaking, however, R and E are nothing more than the parameters of a simple model (see Figure 54-16) that has been forced to match the measured signals as well as possible. The usefulness of R and E as reflections of physiology thus lies in the degree to which the single-compartment linear model accurately describes the behavior of the system under study. In many cases, this accuracy is acceptable, such as the example shown in Figure 54-18 of a patient with chronic obstructive lung disease being mechanically ventilated in the intensive care unit. When the single-compartment linear model does describe a set of respiratory data to an acceptable degree of accuracy, one is then faced with the task of assigning a physiologic interpretation to the values of R and E it provides. It might seem obvious, for example, that R would correspond to the flow resistance of the airway tree. However, this turns out not to be the case, at least not entirely. Studies with the alveolar capsule in dogs and other animals have allowed total lung resistance (ie, R) to be partitioned into airway resistance and tissue resistance.11,32 The latter has been shown to depend greatly on the frequency at which the lungs are oscillated and, at normal breathing frequencies, may constitute the great majority of R.13 It is not until frequency gets well above the range of normal breathing (above about 2 Hz) that the tissue component of R decreases to the point where R is a good reflection of airway resistance alone.3335 In the intact animal, R also contains a significant contribution from the chest wall.36 Nonlinear Single-Compartment Models There are situations in which the single-compartment linear model does not describe a set of respiratory data with acceptable accuracy. The model must then be replaced by a more realistic (and invariably more complex) model. An example typically occurs when the volume excursions of the lungs become large or when the stiffness of the lung tissue increases in

-2 1.0

0.5

0.0 30 20 10 0 0 1 Time (s) 2 data model fit

FIGURE 54-18 Example of pressure, flow, and volume data over a single breath from a mechanical ventilated patient in the intensive care unit. The lower panel also shows the fit to pressure produced by Equation 54-8.

certain diseases. In this case, one often finds that the dynamic elastic behavior of the tissues is significantly better described by a curvilinear function of volume rather than a straight line, as in the linear model. For example, it has been shown in both humans37 and animals38 that a nonlinear model with the equation (t) E1V(t) E2V2(t) P0 P(t) RV (54-9)

sometimes fits the data significantly better than the linear model above (see Equation 54-8). The nonlinear model is structurally the same as the linear model in that it still has only a single compartment being ventilated through a single airway. The difference is that the elastic properties of the tissues surrounding the compartment are nonlinear. Similarly, the linear resistance term in Equation 54-8 can be replaced by two terms representing a flow-dependent resistance as originally proposed by Rohrer.38 Which of these models should be used to describe a given set of respiratory data can be decided using, for example, the F-ratio test applied to the mean squared differences between measured P and P predicted by the various models.38 Another use of the single-compartment model of respiratory mechanics arises when the stiffness of the lung or respiratory system is assessed from the quasistatic pressurevolume (P-V) curve. The P-V curve is obtained by inflating

Measurement Techniques in Respiratory Mechanics

633

and deflating the lungs, either continuously or in steps, slowly enough that the resistive pressure drop across the airways can be neglected. The result is a relationship that embodies the elastic properties of the pulmonary or respiratory tissues, viewed as a single compartment. The model invoked to account for the P-V curve is thus again a singlecompartment model, but now it is nonlinear because the elastic recoil pressure inside the compartment increases disproportionately as volume approaches total lung capacity. A commonly used equation for describing the descending limb of the P-V curve is the exponential expression39 V A BeKP (54-10)

1.0

0.8 Volume (ml) Pressure (cmH2O)

0.6

0.4

0.2

where A, B, and K are constants chosen to make the right side of the equation match the left side as closely as possible. The ascending limb of the P-V curve lies to the right of the descending limb (Figure 54-19), reflecting a phenomenon known as hysteresis. The amount of hysteresis depends on the volume range over which V is cycled and is caused by a number of factors. One of the most important is recruitment of closed airspaces during inspiration that remain open during expiration. Hysteresis may become markedly enhanced in pathologies such as acute lung injury.40

0.0 40

30

20

10

METHODS BASED

ON

MULTICOMPARTMENT MODELS

The single-compartment linear model (see above) generally describes respiratory pressure-flow data well when volume excursions are modest and the volume oscillations are concentrated around a single frequency, such as pertains, for example, during normal breathing or mechanical ventilation. However, the values of R and E obtained using this model vary with frequency. In particular, R decreases markedly as frequency is increased over the range of normal breathing, whereas E correspondingly increases. The main reason for this frequency dependence of R and E in normal lungs is the fact that the respiratory tissues are viscoelastic, that is, they exert a recoil pressure that is a function not only of volume but also of volume history.41 In diseased lungs, additional variation of R and E with frequency may be caused by regional variations in mechanical function throughout the lung, leading to transient redistribution of gas as the lungs are dynamically inflated and deflated.42,43 In any case, the single-compartment linear model no longer suffices as a description of pulmonary or respiratory mechanics when multiple frequencies are involved. Instead, we need to invoke models featuring two or more compartments to account for regional differences in mechanical function throughout the lung.44 Interrupter Technique A technique for assessing lung function that was first introduced nearly a century ago involves the rapid interruption of airflow at the airway opening, while pressure just behind the point of interruption is measured. Initially, it was thought that this maneuver would simply obliterate any resistive pressure drop across the airways, so that the observed sudden change in pressure would reflect Raw. However, work over the past two decades has shown that the sudden change in pressure occurring with

-10 0 5 Time (s)


B

10

15

0.8

0.6

Volume (l)

0.4

0.2

0.0 0 10 20 30

Pressure (cmH2O)

FIGURE 54-19 A, Pressure and volume data obtained from a mouse during stepwise inflation and deflation of the lungs. B, Ascending and descending limbs of the quasistatic pressure-volume (P-V) curve derived from the plateaus of the data in Figure 54-19A.

interruption of flow is accompanied by some rapid damped oscillations and a subsequent further transient change in pressure to a stable plateau (Figure 54-20). The oscillations are mainly owing to ringing of the gas in the central airways,45 whereas the secondary slow pressure change is due

634

Clinical Respiratory Physiology

Damped high-frequency oscillations Pressure (arbitrary untis)

Secondary slower pressure change

Zin is thus nothing more than a description of how R and E vary over a range of frequencies. Zin still requires that the system under study be linear. This assumes that whatever values of R and E are obtained at a particular frequency, their , and V values do not depend on the amplitudes of the P, V signals used to measure them (which, of course, is never precisely the case in practice). Forced Oscillation Technique The measurement of Zin is achieved by the so-called forced oscillation technique in which a flow generator (such as a loudspeaker or piston pump) is used to drive an oscillatory flow into the lungs via the airway opening.50 The frequency range over which the signal oscillates determines the kind of information that will be obtained about respiratory mechanical function. At frequencies below about 2 Hz, much of Zin is determined by the rheologic properties of the tissues, as well as regional mechanical heterogeneities throughout the lung should they exist. Regional heterogeneities can affect the shape of Zin above 2 Hz as well.42,43 At frequencies of hundreds of Hz one obtains information about the acoustic characteristics of the airways.51 Whatever the frequency range, the interpretation of Zin in physiologic terms requires some kind of model of the system under investigation. For example, normal respiratory or pulmonary Zin is described very accurately below about 20 Hz by a model consisting of a uniformly ventilated compartment surrounded by viscoelastic tissue. The compartment is served by a single airway having a newtonian resistance RN, whereas the viscoelastic tissue has an impedance with real and imaginary parts that both decrease hyperbolically with f. The equation for the impedance of this construct, which is frequently referred to as the constantphase model, is35 Zin ( f ) = R N + i2fI + Gti iHti (2f ) (54-13)

Initial rapid pressure change

0.0 Point of flow interruption

0.5 Time (s)

1.0

FIGURE 54-20 Schematic representation of airway opening pressure recorded during interruption of expiratory flow. An initial almost instantaneous jump in pressure is accompanied by rapid damped oscillations, which are followed by a slower further pressure change.

to the viscoelastic properties of the respiratory tissues when the lung is normal12 and may be accentuated by gas redistribution in pathologic situations.46 Interpreting the initial rapid and secondary slower pressure changes has been done on the basis of two-compartment models of respiratory mechanics.44,47,48 The interrupter technique is currently gaining interest among pediatricians,49 who face a particular challenge in trying to assess lung function in young children and infants unable to perform the voluntary maneuvers necessary to generate forced expired flows. However, the interruption of flow is merely a specialized kind of flow perturbation. Understanding the information obtained by applying general flow perturbations to the lungs is best done in the context of the forced oscillation technique and impedance. Input Impedance The frequency dependence of R and E has led respiratory researchers to move to a more global assessment of mechanics based on a quantity known as input impedance (Zin). Zin can be determined over a range of frequencies signal that contains by subjecting the lungs to an oscillatory V multiple frequencies. Zin is then determined by taking the ratio . of the Fourier transform of P to the Fourier transform of V This yields a complex function of frequency Zin(f ) R(f ) iX(f ) (54-11)

where the parameter is related to the two tissue parameters via the relation

H 2 tan 1 ti Gti

(54-14)

with a real part R(f) and an imaginary part X(f), where i = 1. The value of R at each value of f is equal to the resistance of an equivalent single-compartment linear model, so the R(f) is called the resistance. X(f) is called the reactance and at each f is related to the elastance of the equivalent single-compartment model by X( f ) = E( f ) 2f (54-12)

Gti characterizes the viscous dissipation of energy in the tissues and so is related (but not equivalent to) tissue resistance. Hti characterizes the storage of elastic energy in the tissues and is closely related to E. I is an inertance reflecting the mass of the gas in the central airways. The four parameters RN, I, Gti, and Hti together account for the entire frequency spectrum of Zin below 20 Hz in a convenient and compact way. They also allow Zin to be partitioned into a component pertaining to the airways (ie, RN and I) and a component pertaining to the lung periphery (ie, Gti and Hti). Figure 54-21 shows an example of the fit to Zin provided by Equation 54-13. Other Respiratory Impedances The calculation of respiratory impedance is not limited to Zin obtained from P at the airway opening. Any relevant pressure and and V flow signals will do, although the impedance obtained will

Measurement Techniques in Respiratory Mechanics


2 0 Impedance (cmH2O.s/m) -2 -4 -6 -8 -10 -12 -14 0 5 10 Frequency (Hz) 15 20 Real part Imaginary part Model fit

635

FIGURE 54-21 Fit of constant-phase model (Equation 54-13) to Zin data from a mouse.

be different in each case. For example, if pressure is oscillated around the body surface while flow is measured at the mouth, the impedance obtained is known as transfer impedance (Ztr). Ztr has been used in numerous studies in both animals52 and humans.5355 Although in principle Ztr should give similar information about lung function to Zin, its measurement may be associated with some practical advantages. First, when the distal airways of the lung become significantly constricted, the flow oscillations applied at the mouth to measure Zin may become shunted to a large degree into the central airways that have a finite compliance.43 If the amount of flow reaching the lungs becomes small to the point that it approaches the noise level of flow measurements at the mouth, then one effectively loses the ability to probe the lung periphery. In contrast, when pressure oscillations are applied at the body surface, flow is driven from the lung periphery toward the trachea, and shunting into the central airway compartment is minimized. This means that all flow measured at the mouth comes from the lung periphery, giving Ztr a signal-to-noise advantage over Zin for the investigation of severely constricted lungs. Another recently developed variant of respiratory impedance uses the heart as the oscillatory source, producing what has been termed output impedance.56 When a subject relaxes with an open glottis, the beating heart perturbs the lungs sufficiently to generate a small flow that can be measured at the mouth. When the airway opening is occluded, corresponding pressure oscillations are measured just behind the point of occlusion. The impedance determined from these two signals was found to have an essentially zero imaginary part and a real part that corresponded closely to the resistance of the conducting airways, as measured by Zin.56 Yet a further type of respiratory impedance has been obtained by applying forced oscillations in flow to the lungs of dogs through an alveolar capsule.57 The resulting alveolar input impedance (ZA) between 26 and 200 Hz

was found to be well described by a simple model consisting of a subpleural alveolar compartment connecting to the rest of the lungs by a terminal airway compartment. The magnitude of ZA was found, on the basis of an anatomically accurate computer model of the dog lung,58 to correspond to that expected of a single lung acinus. When ZA was followed during the development of bronchoconstriction, the response of the lung periphery was demonstrated to be extremely heterogeneous both spatially and temporally.59,60 The information obtained about lung mechanical function thus depends on the site at which flow perturbations are applied and where pressures and flows are measured. It also depends to a great extent on the nature of the flow perturbations themselves. Measuring lung mechanics during normal breathing in a conscious subject using an esophageal balloon has the advantage of allowing the subject to remain in a reasonably natural state. However, it suffers from the disadvantage that the subject is free to choose the breathing pattern, which may change with an intervention. This itself will affect the measurement of mechanics regardless of any true changes in the intrinsic mechanical properties of the airways or tissues. Thus, one faces a trade-off between minimizing the interventional nature of the measurement and controlling for confounding variables. This situation has been likened to the uncertainty principle of quantum mechanics.61

SUMMARY
The quantitative study of respiratory mechanics involves measurement at two levels, namely the measurement of the raw signals that carry the mechanical information and the measurement of key physiologic parameters that embody this information. Measurement of signals requires the use of transducers for recording pressure, flow, and volume and computers for capturing and storing the data. Measurement of physiologic parameters involves matching the measured signals to a suitable mathematical model of respiratory mechanics. There are many decisions to be made in achieving these ends, such as which type of transducer to use, how fast to sample the signals, what resolution of analog-to-digital converter is required, what kind of perturbation should be applied to the respiratory system, and what mathematical model should be invoked to interpret the data. There is no universally correct decision for any of these issues because the appropriate action to be taken depends on the physiologic questions being addressed. These questions, whatever they are, will determine what mathematical model of respiratory mechanics should be invoked, what frequency response characteristics are required of the transducers, what data sampling rate is needed, and so on. It is therefore important for the respiratory scientist to understand the basics of measurement theory as it applies to both the collection of physiologic signals and their interpretation through mathematical models. Such an understanding minimizes the risk, for example, of being misled by measurement artifact from a transducer that is not up to the task required of it or of being confused by the apparently bizarre behavior of a calculated parameter obtained using a model that does not apply to the situation at hand.

636

Clinical Respiratory Physiology 22. Aliverti A, Dellaca R, Pelosi P, et al. Compartmental analysis of breathing in the supine and prone positions by optoelectronic plethysmography. Ann Biomed Eng 2001; 29:6070. 23. Gold WM. Pulmonary function testing. In: Murray JF, Nadel JA, editors. Textbook of respiratory medicine. Philadelphia (PA): W.B. Saunders; 2000. p. 781882. 24. Lambert RK, Wilson TA, Hyatt RE, Rodarte JR. A computational model for expiratory flow. J Appl Physiol 1982;52:4456. 25. Shore S, Milic-Emili J, Martin JG. Reassessment of body plethysmographic technique for the measurement of thoracic gas volume in asthmatics. Am Rev Respir Dis 1982;126:51520. 26. Drorbaugh J, Fenn W. A barometric method for measuring ventilation in newborn infants. Pediatrics 1955;16:817. 27. Hamelmann E, Schwarze J, Takeda K, et al. Noninvasive measurement of airway responsiveness in allergic mice using barometric plethysmography. Am J Respir Crit Care Med 1997;156:76675. 28. Epstein RA, Epstein MA, Haddad GG, Mellins RB. Practical implementation of the barometric method for measurement of tidal volume. J Appl Physiol 1980;49:110715. 29. Lundblad LK, Irvin CG, Adler A, Bates JH. A reevaluation of the validity of unrestrained plethysmography in mice. J Appl Physiol 2002;93:1198207. 30. Finotto S, Neurath MF, Glickman JN, et al. Development of spontaneous airway changes consistent with human asthma in mice lacking T-bet. Science 2002;295:3368. 31. Chong BT, Agrawal DK, Romero FA, Townley RG. Measurement of bronchoconstriction using whole-body plethysmograph: comparison of freely moving versus restrained guinea pigs. J Pharmacol Toxicol Methods 1998;39:1638. 32. Fredberg JJ, Ingram RH Jr, Castile RG, et al. Nonhomogeneity of lung response to inhaled histamine assessed with alveolar capsules. J Appl Physiol 1985;58:191422. 33. Sato J, Davey BL, Shardonofsky F, Bates JH. Low-frequency respiratory system resistance in the normal dog during mechanical ventilation. J Appl Physiol 1991;70:153643. 34. Hantos Z, Daroczy B, Suki B, Nagy S. Low-frequency respiratory mechanical impedance in the rat. J Appl Physiol 1987; 63:3643. 35. Hantos Z, Daroczy B, Suki B, et al. Input impedance and peripheral inhomogeneity of dog lungs. J Appl Physiol 1992;72:16878. 36. Bates JH, Abe T, Romero PV, Sato J. Measurement of alveolar pressure in closed-chest dogs during flow interruption. J Appl Physiol 1989;67:48892. 37. Bersten AD. Measurement of overinflation by multiple linear regression analysis in patients with acute lung injury. Eur Respir J 1998;12:52632. 38. Wagers S, Lundblad L, Moriya HT, et al CG. Nonlinearity of respiratory mechanics during bronchoconstriction in mice with airway inflammation. J Appl Physiol 2002;92:18027. 39. Salazar E, Knowles JH. An analysis of pressure-volume characteristics of the lungs. J Appl Physiol 1964;19:97104. 40. Hickling KG. The pressure-volume curve is greatly modified by recruitment. A mathematical model of ARDS lungs. Am J Respir Crit Care Med 1998;158:194202. 41. Bates JH, Brown KA, Kochi T. Respiratory mechanics in the normal dog determined by expiratory flow interruption. J Appl Physiol 1989;67:227685. 42. Kaczka DW, Ingenito EP, Israel E, Lutchen KR. Airway and lung tissue mechanics in asthma. Effects of albuterol. Am J Respir Crit Care Med 1999;159:16978. 43. Lutchen KR, Greenstein JL, Suki B. How inhomogeneities and airway walls affect frequency dependence and separation of airway and tissue properties. J Appl Physiol 1996;80: 1696707.

REFERENCES
1. Jackson AC, Vinegar A. A technique for measuring frequency response of pressure, volume, and flow transducers. J Appl Physiol 1979;47:4627. 2. Duvivier C, Rotger M, Felicio da Silva J, et al. Static and dynamic performances of variable reluctance and piezoresistive pressure transducers for forced oscillation measurements. Eur Respir J 1991;1:14650. 3. Bates JHT, Sly PD, Sato J, et al. Correcting for the Bernoulli effect in lateral pressure measurements. Pediatr Pulmonol 1992;12:2516. 4. Navalesi P, Hernandez P, Laporta D, et al. Influence of site of tracheal pressure measurement on in situ estimation of endotracheal tube resistance. J Appl Physiol 1994;77:2899906. 5. Couture JG, Chartrand D, Gagner M, Bellemare F. Diaphragmatic and abdominal muscle activity after endoscopic cholecystectomy. Anesth Analg 1994;78:7339. 6. Baydur A, Behrakis PK, Zin WA, et al. A simple method for assessing the validity of the esophageal balloon technique. Am Rev Respir Dis 1982;126:78891. 7. Dechman G, Sato J, Bates JHT. Factors affecting the accuracy of esophageal balloon measurement of pleural pressure in dogs. J Appl Physiol 1992;72:3838. 8. Peslin R, Navajas D, Rotger M, Farre R. Validity of the esophageal balloon technique at high frequencies. J Appl Physiol 1993;74:103944. 9. Panizza JA. Comparison of balloon and transducer catheters for estimating lung elasticity. J Appl Physiol 1992;72:2315. 10. Wang CG, DiMaria G, Bates JH, et al. Methacholine-induced airway reactivity of inbred rats. J Appl Physiol 1986;61:21805. 11. Fredberg JJ, Keefe DH, Glass GM, et al. Alveolar pressure nonhomogeneity during small-amplitude high-frequency oscillation. J Appl Physiol 1984;57:788800. 12. Bates JH, Ludwig MS, Sly PD, et al. Interrupter resistance elucidated by alveolar pressure measurement in openchest normal dogs. J Appl Physiol 1988;65:40814. 13. Ludwig MS, Dreshaj I, Solway J, et al. Partitioning of pulmonary resistance during constriction in the dog: effects of volume history. J Appl Physiol 1987;62:80715. 14. Ludwig MS, Romero PV, Bates JH. A comparison of the doseresponse behavior of canine airways and parenchyma. J Appl Physiol 1989;67:12205. 15. Tomioka S, Bates JH, Irvin CG. Airway and tissue mechanics in a murine model of asthma: alveolar capsule vs. forced oscillations. J Appl Physiol 2002;93:26370. 16. Renzi PE, Giurdanella CA, Jackson AC. Improved frequency response of pneumotachometers by digital compensation. J Appl Physiol 1990;68:3826. 17. Farre R, Peslin R, Navajas D, et al. Analysis of the dynamic characteristics of pressure transducers for studying respiratory mechanics at high frequencies. Med Biol Eng Comput 1989;27:5316. 18. Schuessler TF, Maksym GN, Bates JHT. Estimating tracheal flow in small animals. Proceedings of the 15th Annual International Meeting of the I.E.E.E. Engineering in Medicine and Biology Society; 1993 October 2831; San Diego, California. p. 5601. 19. Schibler A, Hall GL, Businger F, et al. Measurement of lung volume and ventilation distribution with an ultrasonic flow meter in healthy infants. Eur Respir J 2002;20:9128. 20. Clary AL, Fouke JM. Fast-responding automated airway temperature probe. Med Biol Eng Comput 1991;29:5014. 21. Cohen KP, Ladd WM, Beams DM, et al. Comparison of impedance and inductance ventilation sensors on adults during breathing, motion, and simulated airway obstruction. IEEE Trans Biomed Eng 1997;44:55566.

Measurement Techniques in Respiratory Mechanics 44. Similowski T, Bates JHT. Two compartment modelling of respiratory system mechanics at low frequencies: gas redistribution or tissue rheology? Eur Respir J 1991;4:3538. 45. Romero PV, Sato J, Shardonofsky F, Bates JH. High-frequency characteristics of respiratory mechanics determined by flow interruption. J Appl Physiol 1990;69:16828. 46. Ludwig MS, Romero PV, Sly PD, et al. Interpretation of interrupter resistance after histamine-induced constriction in the dog. J Appl Physiol 1990;68:16516. 47. Bates JH, Baconnier P, Milic-Emili J. A theoretical analysis of interrupter technique for measuring respiratory mechanics. J Appl Physiol 1988;64:220414. 48. Bates JH, Rossi A, Milic-Emili J. Analysis of the behavior of the respiratory system with constant inspiratory flow. J Appl Physiol 1985;58:18408. 49. Frey U, Silverman M, Kraemer R, Jackson AC. High-frequency respiratory input impedance measurements in infants assessed by the high speed interrupter technique. Eur Respir J 1998;12:14858. 50. MacLeod D, Birch M. Respiratory input impedance measurement: forced oscillation methods. Med Biol Eng Comput 2001;39:50516. 51. Dorkin HL, Lutchen KR, Jackson AC. Human respiratory input impedance from 4 to 200 Hz: physiological and modeling considerations. J Appl Physiol 1988;64:82331. 52. Sobh JF, Lilly CM, Drazen JM, Jackson AC. Respiratory transfer impedance between 8 and 384 Hz in guinea pigs before and after bronchial challenge. J Appl Physiol 1997;82:17281.

637

53. Marchal F, Bouaziz N, Baeyert C, et al. Separation of airway and tissue properties by transfer respiratory impedance and thoracic gas volume in reversible airway obstruction. Eur Respir J 1996;9:25361. 54. Peslin R, Gallina C, Duvivier C. Respiratory transfer impedances with pressure input at the mouth and chest. J Appl Physiol 1986;61:816. 55. Peslin R, Gallina C, Teculescu D, Pham QT. Respiratory input and transfer impedances in children 9-13 years old. Bull Eur Physiopathol Respir 1987;23:10712. 56. Bijaoui E, Baconnier PF, Bates JH. Mechanical output impedance of the lung determined from cardiogenic oscillations. J Appl Physiol 2001;91:85965. 57. Davey BL, Bates JH. Regional lung impedance from forced oscillations through alveolar capsules. Respir Physiol 1993; 91:16582. 58. Mishima M, Balassy Z, Bates JH. Assessment of local lung impedance by the alveolar capsule oscillator in dogs: a model analysis. J Appl Physiol 1996;80:116572. 59. Balassy Z, Mishima M, Bates JH. Changes in regional lung impedance after intravenous histamine bolus in dogs: effects of lung volume. J Appl Physiol 1995;78:87580. 60. Mishima M, Balassy Z, Bates JH. Acute pulmonary response to intravenous histamine using forced oscillations through alveolar capsules in dogs. J Appl Physiol 1994;77:21408. 61. Bates JH, Irvin CG. Measuring lung function in mice: the phenotyping uncertainty principle. J Appl Physiol 2003; 94:1297306.

You might also like