Laser-based absorption spectroscopy as a technique for rapid in-line analysis of respired gas concentrations of O₂ and CO₂

Abstract

The use of sidestream analyzers for respired gas analysis is almost universal. However, they are not ideal for measurements of respiratory gas exchange because the analyses are both temporally dissociated from measurements of respiratory flow and also not generally conducted under the same physical conditions. This study explores the possibility of constructing an all optical, fast response, in-line breath analyzer for oxygen and carbon dioxide. Using direct absorption spectroscopy with a diode laser operating at a wavelength near 2 μm, measurements of expired carbon dioxide concentrations were obtained with an absolute limit of detection of 0.04% at a time resolution of 10 ms. Simultaneously, cavity enhanced absorption spectroscopy at a wavelength near 760 nm was employed to obtain measurements of expired oxygen concentrations with an absolute limit of detection of 0.26% at a time resolution of 10 ms. We conclude that laser-based absorption spectroscopy is a promising technology for in-line analysis of respired carbon dioxide and oxygen concentrations.

the 1950s and 60s saw the advent of rapid analysis of human breath, where a sample of gas from the respired stream could be drawn and analyzed continuously for CO₂ and O₂. Generally, the analyzers for CO₂ were based on infrared absorption spectroscopy, and those for O₂ were based either on the paramagnetic effect or on the employment of fuel cells (4). These techniques have remained the mainstream approach for the analysis of CO₂ and O₂ in respired gas and for many applications have relatively few drawbacks. However, the employment of a “sidestream” flow of gas along a sampling catheter introduces a variable delay into the analysis and also limits the time resolution of the measurement. In turn, this limits the ability to time align the analyses of CO₂ and O₂ with instantaneous measurements of respiratory flow. Such alignment is required for accurate breath-by-breath determinations of CO₂ and O₂ exchange by algorithms that integrate the flow of CO₂ and O₂ along the airway (6). Although respired gas analysis by mass spectrometry can partially overcome this problem (1–2, 7, 9), the cost, complexity, and maintenance of such instruments render them unsuitable for many applications.

The present report is of a “proof-of-concept” study to show that laser absorption spectroscopy can be used to determine CO₂ and O₂ directly within the respired gas stream and thus avoid the variable delays and relatively slow response speeds of sidestream analyzers. The primary motivation behind this is to improve technology for measuring breath-to-breath gas exchange, particularly in situations where the inspired gas is of unknown and possibly time-varying composition. Examples of such applications include patients receiving oxygen therapy or undergoing anesthesia. The absorption technique involves tuning a continuous wave diode laser so that the wavelength of light can be varied repeatedly across a small region of the spectrum that contains an absorption line that is totally specific for either CO₂ or O₂. The strength of the absorption then yields a measure of the concentration of the species in the respired gas. For measuring CO₂, transitions at wavelengths ∼2 μm were chosen. The absorption cross sections are strong enough in this region to allow the relatively simple technique of single-pass absorption spectroscopy to provide sufficient sensitivity. The measurement of O₂ presents a different challenge as there are no conveniently placed strong transitions in the O₂ spectrum. Measurements were therefore carried out on weaker transitions ∼760 nm using the more complex, but more sensitive, technique of cavity-enhanced absorption spectroscopy (CEAS). There is, however, an intrinsic source of measurement noise associated with the use of optical cavities, and a particular objective of this study is to determine whether it would be possible to obtain sufficient sensitivity in the measurement of O₂ while at the same time maintaining a rapid dynamic response for the instrument.

METHODS

Absorption spectroscopy.

The measurements made in this work rely on techniques based on absorption spectroscopy, and a brief overview of the theory is presented in this section. Traditional absorption spectroscopy measures the reduction in radiation intensity as it passes through an absorbing medium and correlates this to the absolute concentration (i.e., molecules per unit volume) of absorber present. The intensity of light at the optical frequency ν, that is measured while tuning over an absorption feature, depends on the optical pathlength through the sample, the absorption probability (in the form of the absorption cross section) and the quantity of absorbing species present through the Beer-Lambert law:

$$I\left(v\right)=I_0\left(v\right)e^{-\sigma \left(v\right)\left[N\right]l}$$ (1)

where I(ν) and I₀(ν) are the light intensities that reach the detector with and without an absorber present, σ(ν) is the absorption cross section for a transition at frequency ν, [N] is the concentration of the absorbing species, and l is the pathlength through the sample. While absorption spectroscopy is attractive in that it allows measurement of absolute number densities, it is an inherently low sensitivity technique reliant on measuring a small change on a large background, which is often affected by laser intensity fluctuations and detector noise. CEAS is a more sensitive variant of absorption spectroscopy, as it enhances the effective pathlength through the absorbing medium by trapping the light between two highly reflective mirrors that constitute an optical cavity. I(v) is determined once again by measuring the light that is transmitted through the exit mirror as the laser wavelength is scanned over an absorption feature, and the concentration of absorber present can now be determined using a variation of Eq. 1 (5):

$$\frac{I_o\left(v\right)-I\left(v\right)}{I\left(v\right)}=\frac{\sigma \left(v\right)\left[N\right]l}{1-R}$$ (2)

where R is the geometric mean of the reflectivity of the cavity mirrors. For small absorption in a non-cavity experiment Eq. 1 can be transformed into Eq. 2 with R = 0. It can thus be seen that for CEAS the term (1-R)⁻¹ leads to an effective pathlength enhancement, for example, of a factor of 10⁴ for a reflectivity of R = 0.9999. To obtain an absolute value of the concentration of absorber from Eq. 2 the value of R needs to be known, and this can be achieved by calibrating the system with a gas sample containing O₂ at a known concentration. Generally multiple scans over the absorption feature are needed to enhance the signal-to-noise ratio. A more detailed description of absorption spectroscopy and CEAS, including the theory behind the calculations used in this work can be found elsewhere (3).

The specific transitions used in this experiment were chosen to optimize the level of absorption, i.e., those with the largest cross sections. The values of these cross sections for CO₂ and O₂ have been comprehensively tabulated in the HITRAN molecular spectroscopy database (8). The transitions were also chosen so that they were isolated from other respired gas absorption features that would complicate the analysis. A summary of the transition details is given in Table 1, where the peak cross sections σ(ν) are given against a background of air (21% O₂, 79% N₂) at 296 K and at standard atmospheric pressure. The line assignments are given in the appropriate spectroscopic nomenclature for the CO₂ vibrational and O₂ electronic transitions. We intended to measure the absorption over a pathlength of 4 cm in the respired gas flow. It can be calculated from these data that, for CO₂ at a typical 4% level in respired air, a fractional absorption of 2% is achieved in a 4 cm single path. This measurement is straightforward. The peak absorption cross-section for O₂ is much lower than that for CO₂, and a 21% molecular oxygen concentration would yield a single-pass fractional absorption of only 0.1%. To measure this level of absorption requires the more sensitive CEAS technique, particularly when a precision of the measured absorption of better than 1% is required over a time scale of ∼10 ms.

Table 1.

Summary of spectroscopic information for transitions chosen for measurement of CO₂ and O₂ absorption

	CO2	O2
Line position (in vacuo) (nm)	2003.5	760.445
Peak cross-section at 296 K (cm2molecule−1)	5.32 × 10−21	4.94 × 10−23
Line assignment	R18 (ΔJ, J″)	R11Q12 (ΔN N″ ΔJ J″)
Transition	(2ν1+ν3)	b 1Σg+ − X 3Σg−

Data and notation are from ref (8).

Temperature, pressure, and background gas composition all affect the shape of the transition. In addition, temperature, but not pressure or background gas composition, affects the absorption cross section by altering the proportion of molecules in the particular rotational level to absorb the radiation. These properties can be characterized through parameters that are specific to each particular transition.

Technical description.

A schematic of the overall experimental system is given in Fig. 1. To measure both gases simultaneously, a single measurement chamber is required where both the 760 nm and 2 μm lasers pass through the same part of the breath sample. To achieve this, the two lasers are counterpropagated through a 4-cm-long optical cavity consisting of two concave mirrors with radii of curvature of 1,500 mm (Layertec) that sits perpendicular to the gas flow. Since the mirrors are only highly reflective over a relatively small region around 760 nm, radiation of this wavelength will be reflected to give the enhancement required for O₂ detection, whereas the radiation at 2 μm is simply transmitted. The mirrors are housed in an aluminium flow tube to allow gentle warming of the measurement cell, preventing condensation on the mirrors. At the input and output of the cell, there are sections of heated aluminium baffles keeping the gas temperature inside the measurement cell constant. The ends of the measurement cell have been adapted to hold standard tubing and bacterial filters (Intersurgical).

Fig. 1.

Schematic of overall experimental system. Gas flow is shown in green. The path for radiation at 760 nm (O₂ measurement) is shown in red. The path for radiation at 2 μm (CO₂ measurement) is shown in blue. DFB, distributed feedback diode laser (for measurement of O₂); VCSEL, vertical cavity surface emitting laser (for measurement of CO₂); APD, Si avalanche photodiode (for detection of radiation at 760 nm). The 2 μm radiation is detected with a biased InGaAs photodiode. Radiation at 760 nm passes through an optical isolator to prevent feedback to the DFB.

The laser used for O₂ detection is a fiber-coupled distributed feedback (DFB) diode centered at 760 nm (Eagleyard), the radiation from which is passed through an optical isolator (Leysop) to minimize potentially damaging reflections from feeding back into the laser. The temperature and injection current for the laser are varied with a temperature controller (Thorlabs TED 200C) and a laboratory-built current controller. The radiation that exits the optical cavity is focused onto an avalanche photodiode detector (APD; Hamamatsu, model C5460-01). For the CO₂ monitoring, a vertical cavity surface emitting laser (VCSEL; Vertilas) centered at 2 μm is used in combination with a laser diode controller (Thorlabs, model VITC002). This radiation exits the cell and is focused onto a long-wavelength type InGaAs PIN photodiode (Hamamatsu, model G5852-01).

Data acquisition and analysis.

Each laser is scanned over the absorption feature by modulating the laser injection current with a triangular ramp function of appropriate amplitude and frequency to cover the required wavelength range and to give the desired time resolution for consecutive scans. In the particular configuration employed, data were only processed from the rising phase of the triangular waveform. The amplitude and frequency of the ramp are controlled from a separate computer for each gas species. Each computer is equipped with digital-to-analog and analog-to-digital converters (DAQ-2010, National Instruments) and runs a real-time LabVIEW (National Instruments) program operating under LabVIEW's real-time operating system. These computers also acquire and analyze data from the detectors at a sampling frequency of 70 kHz, which yields 350 data points per waveform at a scan frequency of 1 kHz. To improve the signal-to-noise ratio a number of consecutive scans over a feature can be averaged. However, this does reduce the time resolution of measurements, and hence a compromise between sampling speed and noise reduction is required. An optimum of averaging 10 scans at a frequency of 1 kHz was chosen, which resulted in a readout every 10 ms.

Before use, the instrument in its present configuration requires calibration. First nitrogen is passed through the cell to obtain the background intensity spectrum with no absorber present, I₀(ν). Next gas mixtures of known concentration of O₂ and CO₂ are passed through the cell to obtain the intensity spectrum, I(ν), with a known concentration of absorber present. These two sets of calibration data are processed using Eqs. 1 and 2 for CO₂ and O₂, respectively, to give a reference absorption profile for each gas at a known concentration. This process then allows the calculation of a calibration constant that relates the area of a profile to the concentration of a gas. For these calibration steps, since only a single spectrum is required for each gas, several hundred scans can be averaged together to reduce the noise on the spectrum.

Once calibrated, a subject then breathes through the cell and averaged spectra are taken every 10 ms. These spectra are first processed in the same manner as the calibration spectra to obtain the absorption profiles. For each gas species, the magnitude of the absorption profile is then compared with the magnitude of the reference absorption profile using a linear regression algorithm. The scaling factor obtained can then be converted into a gas percentage. The readouts can be calculated and displayed in real time as well as saved for later analysis.

In the present configuration, there was no correction employed for changes in the shape of the transition with temperature or pressure changes (see discussion). Similarly, no allowance was made for the variation in absorption cross section with temperature. However, the particular transitions chosen had relatively small coefficients for the effects of temperature on absorption.

Human experimentation was conducted in accordance with the Declaration of Helsinki and Title 45, U.S. Code of Federal Regulations, and approved by the Oxford Research Ethics Committee.

RESULTS

Figure 2 illustrates the calibration process for both CO₂ and O₂. The two spectra shown in the plots on the left are for when calibration gas is present and for when pure N₂ is present (I(ν) and I₀(ν), respectively). These spectra are then processed as described above to give the reference absorption profiles that are shown on the right in Fig. 2. For CO₂, the calibration gas mixture consisted of 4.9% CO₂ in 90% N₂/5.1% O₂. Measurements were made at ambient pressure and with the cell heated to 313 K, and the peak absorption of 2.5% shown in Fig. 2 is in good agreement with that calculated from the cross-section data of Table 1. In a separate series of experiments, the CO₂ cross section was measured and found to agree well with the value given in the Hitran compilation. The O₂ calibration was taken on a sample of synthetic air (21% O₂, 79% N₂) under the same pressure and temperature conditions. For the oxygen CEAS measurements these data can also be used to calibrate the mirror reflectivity, found to be 0.9981 at 760.445 nm.

Fig. 2.

Example calibration data. Top, CO₂ calibration; bottom, O₂ calibration. Left, spectra obtained in the presence of pure background gas (N₂) and calibration gas (containing fixed quantities of either CO₂ or O₂). Right, absorption profiles calculated from the calibration spectra.

An example of data taken while a subject is breathing through the cell is shown in Fig. 3. The data show the variations in O₂ and CO₂ concentrations with time, which have been calculated by comparison with the calibration spectra shown in Fig. 2. The measurements are taken every 10 ms and data analysis is carried out in real time. The standard deviations, σ, on the traces of Fig. 2 were calculated during the constant O₂ and CO₂ phase of (air breathing) inspiration. From these, we define a detection limit (2σ) for the absolute percentage changes of these gases in air at atmospheric pressure of 0.26% for O₂ and 0.04% for CO₂, well within the target resolution, and achieved with the desired time resolution of 10 ms.

Fig. 3.

Example data recorded from a subject breathing air under resting conditions. Top, O₂ concentration record; bottom, CO₂ concentration record.

DISCUSSION

The primary outcome of this study is the demonstration that laser spectroscopy has sufficient sensitivity to provide rapid in-line measurement of airway CO₂ using a path length of 4 cm and rapid in-line measurement of airway O₂ using a 4 cm optical cavity. It demonstrates that the apparatus is capable of producing high-quality data with relatively low reflectivity mirrors that can easily be maintained in a clinical setting. The absolute detection levels of 0.26% for O₂ and 0.04% for CO₂ compare with values from our respiratory mass spectrometer (Airspec, QP 9000) of ∼0.16% for O₂ and ∼0.05% for CO₂.

The primary motivation for exploring rapid in-line gas analysis is to provide gas analyses that are precisely aligned in time with measurements of respiratory flow, and so improve our ability to calculate metabolic exchange by direct integration of the product of instantaneous concentration and instantaneous flow. Accuracy in this integration becomes especially important in cases where there is significant bidirectional flow, as is the case for oxygen. Here, the difference between the inflow and the outflow of the gas species can be relatively small compared with the magnitude of either the inflow or the outflow on its own. Thus small errors in integration during either inspiration or expiration tend to have large effects on the overall estimate of gas exchange (10). A further problem with accuracy arises from the variation in temperature and water vapor content of the gas through the respiratory cycle. Here, in-line analysis of gas composition has the potential to reduce these errors by ensuring that the gas composition is always determined under the same conditions of temperature and humidity as the gas flow. Furthermore, although temperature and humidity both vary through the respiratory cycle, the opposing effects of these on concentration and flow are such that the product, which is used to calculate the metabolic exchange, is unaffected. For example, any increase in volume (and hence flow) through an increase in temperature or the addition of water vapor will proportionally reduce the molar concentration of the gas species, and the product of the two variables will remain constant.

The experimental system was constructed to explore the concept of using laser absorption spectroscopy for respired gas analysis rather than for direct physiological use, but a number of its features could be further improved. The measurement cell design could be improved to provide greater stability after calibration. The use of higher reflectivity mirrors in the optical cavity could improve the signal-to-noise ratio for oxygen, although the relatively low measured reflectivity of the mirrors in the current system should allow performance to be easily maintained in a clinical setting. We found some degree of thermal management of the cell was important. First, warming the cell was necessary to prevent condensation on the optical surfaces. Second, we found that without any attempt to reduce the temperature change in the cell between inspiration and expiration, high noise levels were observed in the oxygen spectra particularly when the gas flow changed direction. We do not fully understand the mechanism associated with this increase in noise. Third, any change in temperature propagates through to variations in the population of particular rotational levels of the molecules, therefore leading to an effective change in the absorption cross section, although this issue could be addressed with a more sophisticated approach to calibration and data processing in the software.

The findings from this “proof of principle” study support the case for further work to develop a system suitable for routine physiological use. This work would necessarily include compacting the instrument into a measurement cell that can be mounted in standard mouthpieces with all laser radiation coupled in and out of the cell by the use of fiber optic cables. It would also require simultaneous measurements of temperature, flow, and water vapor pressure (which, like CO₂, has strong absorbance cross sections and so measurements should be possible using single-pass laser spectroscopy). In principle the CO₂ and H₂O concentrations can be measured through Eq. 1 without recourse to using a calibration mixture, providing the lineshapes of each of the transitions are accurately known. Lineshapes depend on total pressure, gas composition, and temperature, and predetermined variations with these parameters would be needed to process the data accurately, particularly for non-standard gas mixtures. In the present experiments, we calculate that a 1-K change in temperature around a cell temperature of 313 K would affect the spectral width by 0.15%. A 5% change in O₂ concentration (from 16–21%) would change the O₂ spectral width by <0.1%, and a 5% change in CO₂ (from 0–5%) would change the CO₂ spectral width by 1.8%. Although the specialist lasers and components used in the construction of this laboratory device are relatively expensive, there is no intrinsic reason why any of them should be so once they are required in larger volumes. In conclusion, laser absorption spectroscopy of respired gases is feasible and offers a number of significant advantages over current techniques for respired gas analysis.

GRANTS

This work was supported by a Medical Research Council Discipline Hopping Award and the Oxford Biomedical Research Centre.

DISCLOSURES

Isis Innovation, a wholly owned subsidiary of the University of Oxford, holds a patent for the commercial exploitation of the invention. Six authors have an interest in that patent (B. Cummings, M. L. Hamilton, R. Peverall, G. A. D. Ritchie, G. Hancock, and P. A. Robbins).