Pneumotach accuracy

The first reasonably accurate flow-measuring device was the Fleisch pneumotachograph which was developed in 1925. Originally the Fleisch pneumotach bounced a light beam off a mirror mounted on a diaphragm and from there onto photographic film, all of which made it difficult to use. World War II saw the development of sensitive pressure transducers, amplifiers and recorders and by 1950 the pneumotachograph went totally electronic and began to be commonly used in routine pulmonary research.

The first spirometers that used a flow sensor came onto the marketplace around 1970. Since that time, flow sensors of one kind or another have made steady inroads and now the majority of test systems use flow sensors and there are only a small handful of volume-displacement spirometer systems still being manufactured.

There are a variety of difficulties involved with measuring gas flow rates and this has driven the development of a number of different flow measurement techniques. As well as the pneumotachograph, there are now Pitot tube, hot wire, turbine and ultrasonic flow sensors. Some of these techniques are more linear than others but none of them are perfectly linear.

The pneumotachograph however, is inherently the most linear method for measuring gas flow rates and has been more completely characterized than all other techniques. For this reason it is probably used in pulmonary function equipment more frequently than any of the other type of flow sensor and is also far more likely to be used in research.

Gas flow through a pneumotach is measured from the difference in pressure across a resistance. There are a variety of ways of creating this resistance but there are only two methods that are relatively linear, the Fleisch and the screen (aka Lilly) pneumotach.

The resistance in the Fleisch pneumotach consists of a set of narrow capillary tubes, parallel to the direction of flow.

Fleisch Pneumotach

In the screen pneumotach the resistance is one or more sheets of narrow-mesh screen placed perpendicular to the direction of flow.

Screen Pneumotach 

Although a pneumotach is the most linear flow-sensing device, it is not perfectly linear. When a pneumotach is calibrated against a series of precision flow rates the pressure-flow relationship always shows a slight curve. The greatest difference between this calibration curve and a straight line is usually small (less than 3%) but it is not negligible and must be accounted for.

Pneumotach Calibration Curve

For Fleisch pneumotachs the primary factor that determines pressure-flow relationships is gas viscosity. The difference in pressure across the resistance is described by Poisselle’s equation for flow through a tube:

Poisselle Equation

Since for any particular pneumotach its length and radius are fixed and it would seem that the only variable would be the velocity (flow rate) of the gas. Viscosity, however, is also variable and depends both on gas composition and on gas temperature.

Gas Viscosity vs Temperature

Room air (~21% oxygen, 78% nitrogen, 1% argon, 25 degrees centigrade) has a viscosity of approximately 0.01846 centipoises. Exhaled air (~16% oxygen, 74% nitrogen, 4% carbon dioxide, 6% water vapor, 1% argon, 35 degrees centigrade) has a viscosity of approximately 0.01824 centipoises which is a decrease in 1.1%. Interestingly, although viscosity increases with temperature the addition of CO2 and H2O, both of which have substantially lower viscosities than oxygen and nitrogen, acts to decrease the overall viscosity of exhaled air. This example is only an approximation, however, since exhaled gas composition and viscosity will differ depending on the exhalation maneuver and where the gases are measured.

The Poisselle equation only applies to laminar gas flow. Turbulence occurs when the velocity of gas through a pneumotach rises above a threshold value which is best described by the Reynolds number. This is calculated from:

Reynolds Number Equation

A flow with a Reynolds number below 2300 is laminar, above 4000 turbulent and in-between it is transitional. For turbulence gas density does matter and this is why helium-oxygen mixtures have been used both for research and therapeutic purposes.

The volume-pressure relationship of screen pneumotachs is complicated but seems to depend primarily on the Reynolds number determined by the diameter of the wire mesh and to the ratio between the hole size and wire diameter which in turn means that both gas density and gas viscosity can affect linearity.

The Reynolds number formula also shows that the narrower the diameter of the pneumotach the lower the flow rate at which turbulence occurs. The ATS-ERS standards state that a spirometer should be able to accurately measure flow rates up to +/- 14 liters/second. Theoretically this should determine the optimal size of a pneumotach used for pulmonary function testing but there is a tradeoff involved. In order to measure higher flow rates the diameter of a pneumotach needs to increase but when it does it will have a lower pressure/flow relationship and will be correspondingly less sensitive to low flow rates. Since the maximum flow rates are only rarely achieved the pneumotach for a test system may be selected that is more sensitive and accurate over a lower range of expiratory flows with the expectation that some kind of correction will be made for higher flow rates.

Turbulence is not necessarily bad. When gas flow is laminar the velocity of gas is highest in the center of a pneumotach and lowest at the outer edges.

Laminar Flow Gas VelocityThe differential pressure within a pneumotach is measured at the edges of the pneumotach, not the center and for this reason is lower than would be expected from the average velocity of the gas flow. Turbulent flow tends to have a much “flatter” velocity profile and I suspect that some types of pneumotach (the variable-orifice type in particular) exploit this fact. Even so, all pneumotachs will make a transition from laminar to turbulent flow at some point, and the pressure-flow relationship is somewhat different on either side of the transition. As importantly, turbulent flows are “noisier” and this reduces the precision of flow measurements.

The Fleisch and screen pneumotachs depend on laminar flow and the Reynolds formula assumes that gas stream entering the pneumotach is already laminar. This means that what happens upstream of a pneumotach does matter. Asymmetrical velocity profiles caused by sharp bends, narrowings and interior projections can induce turbulence and so some care must be made in a test system’s architecture to prevent this.

Temperature is probably the most important factor affecting pneumotach accuracy. One reason is that the volume of inhaled air increases as it is warmed to body temperature and becomes saturated with water vapor inside the lung. At a room temperature of 20 degrees centigrade a liter of inhaled air expands by approximately 13%. Some of this expansion is lost during exhalation as heat and water vapor are recovered by the respiratory tract and more is lost between the patient’s mouth and the pneumotach but it still means that a greater volume of air is exhaled than was inhaled. This is usually corrected by the BTPS calculations but it is not as clear to me as I’d like about how the BTPS correction should be performed on exhaled volumes measured by pneumotachs.

The original notion for BTPS correction came from the difference between the temperature in the interior of a volume-displacement spirometer and that of exhaled air. A pneumotach may be at room temperature but compared to a volume displacement spirometer it has negligible thermal mass and does not affect the temperature of exhaled air anywhere nearly as much. I think that the distance the pneumotach is placed from an individual’s mouth has a greater influence on exhaled air temperature than the temperature of the pneumotach itself.

However, even though a pneumotach’s thermal mass is low, successive exhalations will cause its temperature to increase which in turn means that the BTPS correction factor needs to change as well. This change in temperature will be related to the number of spirometry maneuvers performed and how close in time they are performed to each other but is not necessarily predictable.

It is possible to measure exhaled air temperature within a pneumotach but this appears to be done only rarely. To some extent this is understandable because measuring exhaled air temperature is technically difficult and has a set of problems all its own but it does mean that when temperature is not measured the test system software will need to make assumptions of some kind in order to perform the BTPS correction.

Because pneumotachs are more or less at room temperature this means that the water vapor in exhaled air can condense on the resistance element (screen or tubes). Condensation acts to increase the resistance through the pneumotach and at least one study showed that after a series of exhalations expiratory volume can be overestimated up to 7% because of condensation.

It used to be common for pneumotachs to be heated to prevent condensation from occurring but this is now relatively rare. More than one study has shown that a heated pneumotach is more stable not only due to a lack of condensation but because its internal temperature is more stable and less affected by the temperature of exhaled air. Heating a pneumotach requires a fair amount of power which makes it unsuitable to portable applications and although it mitigates a number of temperature-related issues it does not completely abolish them.

The accuracy and linearity of a pneumotach also depends on the accuracy and linearity of the transducer that measures its differential pressure. The ATS-ERS statement on Spirometry states that spirometry equipment should be able to measure flow rates of +/- 14 liters/sec but does not set standards for the minimum expiratory flow rate that should be measurable except when discussing end-of-test (EOT) criteria. There it states that an exhaled volume that changes less than 0.025 L in 1 second meets EOT criteria. This means that a pneumotach needs to differentiate flow rates below 25 ml/sec in order to accurately determine EOT.

The maximum differential pressure range for a pneumotach is usually less than +/- 2 cm H2O and this means that to meet EOT criteria a pressure transducer must be sensitive enough to able to measure a differential pressure of less than 0.0036 cm H2O. The transducer also needs to be linear across the entire pressure range. One study, admittedly from over twenty years ago, showed a number of discrepancies in transducer linearity when actual results were compared to the manufacturer’s specifications although interestingly these tended to be at the highest pressures not the lowest.

Pneumotachographs used to be calibrated using a variable flow source and a precision rotameter (flowmeter). A range of flow rates were applied one at a time and a flow/pressure curve that described the characteristics of the pneumotach was slowly built using this process. The pneumotachs used for pulmonary function testing are now universally calibrated using a 3-liter syringe. Most test systems only require a few strokes of the syringe at different flow rates for a calibration. The flow range used for calibration differs from one system to another, but most probably are in the range of 1 to 5 liters/second.

During a series of syringe strokes the flow/pressure curves are analyzed mathematically and used to generate calibration curves. This process was first described over thirty years ago but at that time 50 strokes of the syringe were needed to develop the flow/pressure curves. This process was updated more recently (2003) but even then at least 10 strokes were required to develop the calibration curves. This doesn’t necessarily mean that the small number of strokes used to calibrate test systems are not sufficient since a relatively large number of strokes may be necessary when calibrating a pneumotach with unknown characteristics. Of more concern, at least conceptually, is the is that calibration is performed with room air at room temperature and the flow rates used are less than the maximum range of the pneumotach.

I started my career with volume-displacement spirometers. They have their own problems involving inertia, back pressure, frequency response, leaks and temperature. In addition they tend to be large, mechanically complex and relatively non-portable. For all these reasons I understand why manufacturers have moved towards systems that use flow sensors instead. The biggest advantage of volume-displacement spirometers however, was that it was relatively easy for anybody to verify their accuracy and linearity, and for this reason I am sorry they are no longer as common as they once were.

For all their simplicity there are numerous factors that can affect the accuracy of gas flow measurements made by a pneumotachograph. I do not think these factors are as widely known and appreciated as they should be. This is partly because this information is often hard to find (in older journals behind paywalls), partly because it is often hard to understand (the math can be very dense and authors frequently don’t write for a non-math audience) and partly because textbooks tend to spend more time explaining how pneumotachs work than on the reasons they don’t.

I am not suggesting that the computerized pneumotach systems we use in pulmonary function testing are inaccurate. These devices are usually developed and validated using computer-controlled motor-driven syringes capable of generating the standard ATS waveforms with physiologically correct gases at body temperature. Nevertheless, numerous assumptions about gas viscosity, laminar flow, temperature and electronic linearity are designed into these systems and I am concerned that we do not know what they are or how well they work. Nor is there any easy way to verify their accuracy since most labs do not have the expertise, the equipment or the time needed to perform advanced verifications.

A large amount of time and effort is put into the development of pulmonary function test systems and manufacturers have every right to consider their engineering and computer algorithms to be proprietary information. Manufacturers shouldn’t be required to disclose this proprietary information but I’d like to see something better than just a paper claim of accuracy. There should be some kind of middle ground where the factors that affect flow measurement accuracy are acknowledged, tested and verified. One place this could be addressed would be in the next set of ATS-ERS statements on pulmonary function testing. The current ATS-ERS standards were released almost a decade ago and set only broad outlines for test system specifications (although admittedly better than what came before). I think it is time for the bar to be raised and for a more precise set of requirements to be adopted.

References:

Brusasco V, Crapo R, Viegi G. ATS/ERS Task Force: Standardisation of lung funcition testing. Standardisation of spirometry. Eur Respir J 2005; 26: 319-338.

Duvivier C, Peslin R, Gallina C. An incremental method to assess the linearity of gas flowmeters: application fo Fleisch pneumotachographs. Eur Respir J 1988; 1:661-665.

Finucane KE, Egan BA, Dawson SV. Linearity and frequency response of pneumotachographs. J Appl Physiol 1972; 32(1): 121-126.

Frye RE, Doty R. A comparison of response characteristics of airflow and pressure transducers commonly used in rhinometry. IEEE Transactions of Biomedical Engineering 1990; 37: 937-944.

Gelfand R, Lambertson CJ, Peterson RE, Slater A. Pneumotachograph for flow and volume measurement in normal and dense atmospheres. J Appl Physiol 1976; 41(1): 120-124.

Hankinson JL, Viola JO, Petsonk EL, Ebeling TR. BTPS Correction for ceramic flow sensor. Chest 1994; 105: 1481-1486.

Johns DP, Pretto JJ, Streeton JA. Measurement of gas viscosity with a Fleisch pneumotachograph. J Appl Physiol 1982; 53(1): 290-293.

Miller MR, Pincock AC. Linearity and temperature control of the Fleisch pneumotachograph. J Appl Physiol 1986; 60(2): 710-715.

Miller MR, Sigsgaard T. Prevention of thermal and condensation errors in pneumotachographic recordings of the maximal forced expiratory maneuver. Eur Respir J 1994; 7: 198-201.

Shephard RA. Pneumotachographic measurement of breathing capacity. Thorax 1955; 10: 258-268.

Tang Y, Turner MJ, Yem JS, Baker AB. Calibration of pneumotachographs using a calibrated syringe. J Appl Physiol 2003; 95: 571-576.

Townsend MC, Hankinson JL, Lindesmith LA, Slivka WA, Stiver G, Ayres GT.  Is my lung function really that good? Flow-type spirometer problems that elevate test results.  Chest 2004; 125: 1902-1909.

Turney SZ, Blumenfeld W. Heated Fleisch pneumotachometer: a calibration procedure. J Appl Physiol 1973; 34(1): 117-121.

Van den Boom G, Van der Star LM, Folgering H, Van Schayck CP, Van Weel C. Volume calibration alone may be misleading. Respiratory Medicine 1999; 93: 643-647.

Yeh MP, Gardner RM, Adams TD, Yanowitz FG. Computerized determination of pneumotachometer characteristics using a calibrated syringe. J Appl Physiol 1982; 53(1): 280-285.

Creative Commons License
PFT Blog by Richard Johnston is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

6 thoughts on “Pneumotach accuracy

  1. Hello !

    Great Blog! I have a question

    I have read many spirometer datasheets that say “no calibration” needed. on probing more, I see that its because they use Turbine based flow measurement principle. Can you help me understand why is a calibration check needed daily for a Pneumotachograph based Spirometer vs Turbine based ?

    thanks!

    Joel

    • Joel –

      First, a differential pressure transducer used to determine the flow through a pneumotach is an analog device and the zero and gain of the amplifier circuitry can drift over time. Second, there tends to be a buildup of condensates and dust on the pneumotach’s resistive element (screen or parallel tubes) and this changes the pressure-flow relationship over time as well.

      I am suspicious however, of any claim that a turbine spirometer does not need calibration. Because of inertia in the turbine vane, turbine sensors are subject to lag. When flow rates are rapidly changing, like during a forced vital capacity, their rate of rotation doesn’t instantaneously reflect the actual flow rate through the device. Software looks at the rate of change in vane rotation and “predicts” what the actual flow rate is. The vane in a turbine can also have condensate and dust buildup and this changes its inertia which in turn will affect turbine lag.

      – Richard

  2. Agreed about a daily “verification” (or calibration check) of either systems that claims you need no “calibration” or systems that you do actually calibrate. When you perform this verification you are testing the “Spirometer system” – Including syringe, transducer, software, computer, and operator.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.