1.

Introduction

The presence of endogenous nitric oxide $\left(\mathrm{NO}\right)$ in exhaled breath of humans and animals was first reported in 1991.¹ Since then, with more than 1000 publications in the field, it is becoming increasingly apparent that measurements of exhaled $\mathrm{NO}$ constitute a new way to monitor the inflammatory status in respiratory disorders, such as asthma and other pulmonary conditions. Exhaled nitric oxide as a measure of inflammation is suggested as providing the best combination of disease evaluation and practical implementation for improved asthma outcomes.² Exhaled nitric oxide has been successfully employed in chronic asthma treatment monitoring to reduce the dose of inhaled corticosteroids, which have serious side effects, without compromising asthma control.³ In treating asthma, which affects 19 million Americans, exhaled $\mathrm{NO}$ may soon be incorporated into routine clinical care.⁴

Exhaled $\mathrm{NO}$ concentrations from the lower respiratory tract exhibit significant expiratory flow rate dependence.⁵ Because of this, exhaled $\mathrm{NO}$ is commonly collected using a single breath maneuver at a constant exhalation flow rate. Exhaled $\mathrm{NO}$ sharply rises and reaches a plateau, which can have a positive, negative, or near-zero slope. The plateau concentration is defined as the average concentration over a 3-sec window in the plateau region.⁶ The plateau level is reported as the exhaled $\mathrm{NO}$ value. Measurements of the $\mathrm{NO}$ plateau at multiple flow rates allow for determination of the origin of $\mathrm{NO}$ (proximal or distal lung) by estimating the flow-independent exchange parameters.⁷

Both offline and online techniques are used to study exhaled $\mathrm{NO}$ . In offline sampling, a portion of the exhaled breath is collected in a suitable reservoir for subsequent analysis. In online sampling, $\mathrm{NO}$ is measured over time during exhalation. Offline sampling has the advantages of remote collection, independence from instrument response time, and more efficient use of the analyzer, because breath from several patients may be rapidly analyzed in succession. Online sampling has the advantages of immediate identification of contamination from gas not derived from the lower airways, near-instantaneous test results, and no error introduced by sample storage and handling. Online $\mathrm{NO}$ analysis is generally preferred for the rapid test result. In both on- and offline collection, breath samples are typically collected at a standardized flow rate using a single exhalation.⁶ Sampling during tidal breath, either spontaneous or paced, is gaining momentum as a noninvasive assessment of lung inflammation in infants and children less than 5 years of age.^{8, 9}

The American Thoracic Society and European Respiratory Society (ATS) jointly published requirements for a nitric oxide analyzer.⁶ According to the requirements, a $\mathrm{NO}$ analyzer should have a sensitivity of $<1\mathrm{ppbv}$ with a response time of $<0.5\sec$ . Plateau $\mathrm{NO}$ measurements during tidal breathing range from 1 to $10\mathrm{ppb}$ and occur over a short time $(<0.5\sec )$ at the end of a tidal exhalation.^{8, 9} Therefore, investigations of $\mathrm{NO}$ measurements during tidal breathing require a $\mathrm{NO}$ sensor with a $<0.2$ -sec response time and an equal or better sensitivity than is required for sampling during a single breath maneuver.

Simultaneous measurement of the exhaled ${\mathrm{CO}}_2$ profile can be used as a marker of end-exhaled air when collecting breath from mechanically ventilated patients and subjects who are breathing normally.¹⁰ A rapid rise in the inspiratory flow rate and a sharp decrease in end-tidal ${\mathrm{CO}}_2$ concentrations can be used to adjust and synchronize the $\mathrm{NO}$ and ${\mathrm{CO}}_2$ waveforms both during recording and data processing after $\mathrm{NO}$ and ${\mathrm{CO}}_2$ measurements.¹¹ Additionally, end-tidal ${\mathrm{CO}}_2$ may be useful as an internal standard to verify correct breath collection and as a normalization procedure to reduce variation in calculating exhaled $\mathrm{NO}$ values, as proposed by Roller ¹² This technique was used to measure exhaled $\mathrm{NO}$ in a diverse population.¹³

2.

Available Technologies for Exhaled Nitric Oxide Detection

Several technologies for exhaled nitric oxide measurement have been reported and are summarized in Table 1 . In electrochemical analysis, a voltage potential is applied across two electrodes. $\mathrm{NO}$ sensing is achieved through oxidation of $\mathrm{NO}$ at the electrode surface and measuring the resulting current between the electrodes. Recently, a portable electrochemical analyzer was introduced¹⁴ and compared to chemiluminescence analyzers described later.^{15, 16} The electrochemical analyzer showed good agreement with the chemiluminescent analyzer in both studies. However, the 5-ppb sensitivity of the electrochemical analyzer prevents its use in determining the flow-independent $\mathrm{NO}$ exchange parameters and for $\mathrm{NO}$ concentration measurement during spontaneous breathing.

Mass spectrometry relies on the nonselective ionization of molecules within a sample, requiring a mass-selective apparatus to enhance selectivity. One selective ionization technique is resonance-enhanced multiphoton ionization (REMPI). In this technique, a target species is photoionized and detected using time-of-flight mass spectrometry (TOF-MS). Recently, a system combining REMPI with TOF-MS techniques was reported capable of breath $\mathrm{NO}$ measurements.¹⁷ Short, Frey, and Berter report sub-ppbv detection sensitivity in less than $1\sec$ . A merit of this technique, as well as the optical techniques discussed later, is that multiple target gas species can be detected. Another advantage is the capability to measure concentrations of isotopomers of $\mathrm{NO}$ , which may be useful in metabolic studies.

The chemiluminescent technique is based on a gas-phase reaction between $\mathrm{NO}$ and ozone $\left({\mathrm{O}}_3\right)$ .¹⁸ Oxygen supplied to the instrument is converted to ozone via a high voltage discharge. $\mathrm{NO}$ reacts with ozone in a reaction chamber at a pressure of 4 to $7\mathrm{torr}$ to produce nitrogen dioxide in the excited state, which releases its energy by emitting light in the visible $(>600\mathrm{nm})$ spectrum. A photomultiplier converts the emitted light to an electrical current. The amount of light produced is proportional to the amount of nitric oxide in the sample. Current clinical and point-of-care chemiluminescent analyzers for exhaled nitric oxide breath analysis are based on instrumention developed primarily by Aerocrine, Incorporated,¹⁹ (Providence, New Jersey) and Sievers, Limited (GE Analytical Instruments, Boulder, Colorado). The chemiluminescent analyzer used in this work (model 280, Sievers) has a sensitivity of $<1\mathrm{ppb}$ . However, the need for frequent calibration and high voltage operation has impacted the wide-scale use of chemiluminescent instruments for exhaled $\mathrm{NO}$ monitoring in clinics.

Optical techniques have been employed for exhaled nitric oxide detection including photoacoustic, laser absorption, and cavity-enhanced spectroscopy. A sensor using photoacoustic spectroscopy was used to detect and quantify $\mathrm{NO}$ , but the reported detection sensitivity was $\sim 500\mathrm{ppb}$ .²⁰ The sensitivity of such a sensor can be improved by use of distributed feedback (DFB)-QC lasers with higher output power, because the photoacoustic signal detected by the microphones is directly proportional to the incident optical power.

Laser absorption spectroscopy allows sensitive, selective, and fast-response $\mathrm{NO}$ concentration measurements. For example, a lead salt laser-based system using a multipass cell has achieved a detection sensitivity of $1.5\mathrm{ppbv}$ with a 4-sec integration time and is commercially available.²¹ More recently, two thermoelectrically cooled continuous-wave (cw) quantum cascade laser (QCL)-based spectrometers, which achieved sub-ppbv detection sensitivities^{22, 23} using a multipass cell, were reported. An innovative QCL-based sensor technology that exploits magnetic field modulation in a Faraday rotation configuration²⁴ has been reported, but so far a minimum $\mathrm{NO}$ detection limit of $41\mathrm{ppbv}$ was observed, which can be improved by using a QC laser targeting a ro-vibrational line in the Q-branch.

Very sensitive trace gas measurements can be performed using cavity ring-down spectroscopy (CRDS). This technique uses a sample cell constructed of two ultra-low-loss dielectric mirrors (with reflectivities of $R>99.99\%$ ), which form a high finesse cavity. A CRDS measurement consists of coupling the laser radiation into the high finesse cavity, rapidly switching off the laser source, and measuring the decay rate of the cavity output radiation. The advantages of CRDS are high sensitivity at sub-ppbv levels^{25, 26} due to a long effective optical path, immunity to laser power fluctuations, and self calibration. Halmer used a cavity ring-down absorption spectrometer based on a continuous-wave $\mathrm{CO}$ laser to achieve a minimum detectable concentration for $\mathrm{NO}$ of $0.8\mathrm{ppb}$ in $1\sec$ .²⁷

Integrated cavity output spectroscopy (ICOS) is an alternative method, which also takes advantage of a high finesse cavity, but is less technically demanding than CRDS, as it does not require mode matching or high speed signal sampling electronics required for precise ring-down event measurements. In ICOS only the total laser intensity exiting the high finesse cavity is recorded and not its time dependence. Hence this method relies on exciting many cavity modes simultaneously.^{25, 28} Additional dithering of one of the cavity mirrors improves the cavity throughput and helps minimizing mode noise in the resulting absorption spectra. Off-axis injection of the laser beam into ICOS cell provides an increase of spectral density of cavity modes and thus further minimizes the mode noise.^{29, 30} Several ICOS-based $\mathrm{NO}$ detection systems have been reported.^{31, 32, 33} A mid-infrared, cw, thermoelectrically cooled, quantum cascade laser^{33, 34, 35} is an ideal spectroscopic source for ICOS-based sensor platforms for medical diagnostics because of its high power $(>0.1\mathrm{W})$ and narrow laser spectral width $(\le 3\mathrm{MHz})$ , necessary for efficient coupling of laser radiation into an ICOS cavity. In this work, we used an off-axis ICOS-based sensor employing a 50-cm-long optical cavity. Details of the performance characteristics of the ICOS-based sensor using a quantum cascade laser operating at $5.47\mathrm{\mu }\mathrm{m}$ $\left(1828{\mathrm{cm}}^{-1}\right)$ were reported previously by the authors.³⁶ In this work, a liquid nitrogen cooled quantum cascade laser operating at $5.22\mathrm{\mu }\mathrm{m}$ $\left(1915{\mathrm{cm}}^{-1}\right)$ was used. We report what we believe to be the first quantum cascade laser-based sensor capable of combined tidal $\mathrm{NO}$ and ${\mathrm{CO}}_2$ concentration measurements.

3.

Experimental Method

4.

Results

4.2.

Offline Nitric Oxide and Carbon Dioxide Comparison

72 breath samples were measured first with the ICOS-based $\mathrm{NO}$ sensor followed by the chemiluminescent analyzer, and 65 breath samples were measured in the reverse order, with several hours lapsing between measurements. For exhaled $\mathrm{NO}$ , no change in agreement of the results obtained with different sensors was noted during the entire period of this study. Offline $\mathrm{NO}$ measurements were performed with a 1.8-sec integration time with both $\mathrm{NO}$ sensors. The best to-date minimum detectable $\mathrm{NO}$ concentration achieved with the ICOS sensor is $0.4\mathrm{ppbv}$ with a 1-sec integration time, as determined by $1\sigma$ of the Voigt fit residual using a 76-ppb $\mathrm{NO}$ calibration gas in ${\mathrm{N}}_2$ as a balance gas. The average difference in $\mathrm{NO}$ concentration measurements between the ICOS and chemiluminescent analyzers (ICOS minus chemiluminescence) was $-0.3\mathrm{ppbv}$ , and the limits of agreement (as $\pm 2\sigma$ ) were $-4.7$ and $4.3\mathrm{ppbv}$ $\mathrm{NO}$ . It was observed that the $\mathrm{NO}$ concentration gradually decreases with time in the collected breath samples. To remove systematic errors, the offline $\mathrm{NO}$ measurements were separated into two groups—those measured with ICOS and then chemiluminescence, and those measured in the reverse order. Figure 1 shows Bland-Altman plots for offline $\mathrm{NO}$ concentration measurements for the sequence of ICOS followed by chemiluminescence [Fig. 1a] and chemiluminescence followed by ICOS [Fig. 1b]. $\mathrm{NO}$ concentration measurements from samples measured by ICOS first had an average difference (ICOS minus chemiluminescence) of $1.1\mathrm{ppb}$ and limits of precision of $\pm \sim 2.9\mathrm{ppb}$ ( $-1.2$ - and 4-ppb $\mathrm{NO}$ ). $\mathrm{NO}$ concentration measurements from samples measured by the chemiluminescent analyzer first had an average difference (ICOS minus chemiluminescence) of $-0.7\mathrm{ppb}$ and limits of precision of $\pm \sim 2.9\mathrm{ppb}$ ( $-3.6$ - and 2.1-ppb $\mathrm{NO}$ ).

Fig. 1

Bland-Altman plots comparing ICOS and chemiluminescence for offline $\mathrm{NO}$ analysis. Each data point represents a pair of measurements plotted as the difference between measurements versus their average. (a) 72 offline $\mathrm{NO}$ samples measured with ICOS then chemiluminescence. Solid line represents the mean difference between values (ICOS minus chemiluminescence) obtained using the two methods (1.1-ppbv $\mathrm{NO}$ ). Dashed lines represent the limits of agreement, $-1.7$ - and 4-ppbv $\mathrm{NO}$ ; SD is standard deviation. (b) 65 offline $\mathrm{NO}$ samples measured with chemiluminescence then ICOS, with a mean of difference of $-0.7\mathrm{ppbv}$ and limits of agreement $-2.1$ - and 3.6-ppbv $\mathrm{NO}$ .

For exhaled ${\mathrm{CO}}_2$ , 28 breath samples were collected during a period of two weeks. Offline ${\mathrm{CO}}_2$ measurements were made using a 1-sec averaging time. The average difference between ICOS and NDIR shown in Fig. 2 was $0.01\%{\mathrm{CO}}_2$ , and the limits of agreement were $-0.12$ and $0.14\%{\mathrm{CO}}_2$ .

Fig. 2

Bland-Altman plot comparing ICOS and NDIR for offline ${\mathrm{CO}}_2$ analysis. Each data point represents a pair of measurements plotted as the difference between measurements versus their average. Solid line represents the mean difference between values obtained using the two methods $(0.01\%{\mathrm{CO}}_2)$ ; dashed lines represent the limits of agreement, $-0.12\%{\mathrm{CO}}_2$ and $0.14\%{\mathrm{CO}}_2$ ; and SD is standard deviation.

4.3.

Online Nitric Oxide and Carbon Dioxide Comparison

Figure 3 shows exhaled $\mathrm{NO}$ at $3\mathrm{l}∕\min$ measured with ICOS with a 2.4-sec averaging time and computation time, and chemiluminescence with a 2.4-sec averaging time. The solid line represents the average of 15 breath measurements with the chemiluminescent analyzer at a 16 samples/sec rate after applying a 50-point Savitzky-Golay smoothing routine with a first-order polynomial. The squares represent the average $\mathrm{NO}$ concentration measured by the chemiluminescent analyzer after a 2.4-sec averaging (wide cap error bars are 1 standard deviation for the 15 chemiluminescent breath measurements). The solid circles connected with a dashed line represent average $\mathrm{NO}$ measurements with ICOS using a 2.4-sec averaging and computation time (narrow cap error bars are 1 standard deviation for the 15 ICOS breath measurements).

Fig. 3

Online nitric oxide concentrations at 3-l/min exhalation as a function of time. 15 exhaled breaths were measured online with (a) ICOS and (b) chemiluminescence each with 2.4-sec averaging/computation time. Solid line represents Sievers data (16 samples/sec) smoothed with a 50 point first-order Savitzky-Golay routine. Squares with large cap error bars represent Sievers data after 2.4-sec averaging; circles with dashed line and small cap error bars represent ICOS data after 2.4-sec averaging and computation; and error bars are $\pm 1$ standard deviation.

The end-tidal ${\mathrm{CO}}_2$ did not vary significantly with exhalation flow rate. The standard deviation of end-tidal ${\mathrm{CO}}_2$ values was equivalent between sensors ( $\sigma =0.2\%{\mathrm{CO}}_2$ for ICOS and NDIR). Figure 4 shows exhaled ${\mathrm{CO}}_2$ at $2\mathrm{l}∕\min$ measured with off axis ICOS (1.8-sec data averaging and computation time) [Fig. 4a] and NDIR (1.8-sec data averaging) [Fig. 4b]. A time delay between the start of the NDIR ${\mathrm{CO}}_2$ waveform and the ICOS waveform was observed for all exhalation flow rates.

Fig. 4

Online $\%{\mathrm{CO}}_2$ at a 2-l/min exhalation flow rate. Five successive exhaled breaths were measured online with (a) ICOS and (b) a nondispersive infrared (NDIR) capnograph each with a 1.8-sec averaging time.

4.4.

Tidal Nitric Oxide and Carbon Dioxide Comparison

An $\mathrm{NO}$ waveform was obtained with both the ICOS and chemiluminescence analyzers. Figure 5 shows exhaled $\mathrm{NO}$ data collected from the ICOS with a 0.33-sec data averaging and computation $(3\mathrm{samples}∕\sec )$ [Fig. 5a] and chemiluminescence with a $16\mathrm{samples}∕\sec$ rate during tidal breathing [Fig. 5b]. The chemiluminescent data were not averaged to preserve rapid changes in $\mathrm{NO}$ over time. For tidal ${\mathrm{CO}}_2$ , the in-line NDIR capnograph detected rapid changes in ${\mathrm{CO}}_2$ , whereas the ICOS sensor had a response delay due to sample cell filling. Figures 6a and 6b show exhaled ${\mathrm{CO}}_2$ data collected from the ICOS (0.15-sec averaging and computation) and NDIR (0.15-sec averaging time), respectively. The NDIR capnograph sensor measured a typical expiratory ${\mathrm{CO}}_2$ waveform⁴¹ consisting of three phases: phase 1, a portion of zero ${\mathrm{CO}}_2$ ; phase 2, a rapid rise in ${\mathrm{CO}}_2$ ; and phase 3, a plateau region with a small positive slope. The ICOS sensor measured a ramp waveform with end-tidal ${\mathrm{CO}}_2$ values in agreement with NDIR. Due to the slower gas delivery system, the standard ${\mathrm{CO}}_2$ phases visible in the waveform acquired by the NDIR capnograph were strongly distorted in the ICOS tidal ${\mathrm{CO}}_2$ waveform.

Fig. 5

Exhaled nitric oxide during normal tidal breathing measured with (a) ICOS (0.33-sec averaging and computation time) and (b) chemiluminescence (0.0625-sec averaging time).

Fig. 6

Exhaled ${\mathrm{CO}}_2$ during normal tidal breathing measured with (a) ICOS and (b) a nondispersive infrared (NDIR) capnograph using a 0.14-sec averaging time. The three phases of a standard ${\mathrm{CO}}_2$ expiratory waveform are indicated on the first NDIR expiratory waveform as roman numerals with dashed lines as boundaries. Phase I is the first portion of exhalation with zero ${\mathrm{CO}}_2$ ; phase II is the rapid rise in ${\mathrm{CO}}_2$ that occurs $\sim 1\sec$ after beginning of exhalation; and phase III is the ${\mathrm{CO}}_2$ plateau region.

Table 1

Performance summary of five exhaled nitric oxide measurement techniques.

Sensor characteristics	ATS required specifications	Chemiluminescence	Electro-chemical	REMPI TOF-MS	Mid-IR multipass cell	ICOS
Sensitivity	$1\mathrm{ppb}$	$<1\mathrm{ppb}$	$5\mathrm{ppb}$	$<1\mathrm{ppb}$	$<1\mathrm{ppb}$	$<1\mathrm{ppb}$
Response time	$<0.5\sec$	$<0.1\sec$	$<2\min$ (total measurement time)	$<1\sec$	$<1\sec$	$<1\sec$
Calibration		Every 14 days or when restarted	Not required	External calibration standard	Not required	Internal

5.

Discussion

The ICOS offline measurements of exhaled $\mathrm{NO}$ and ${\mathrm{CO}}_2$ are in good agreement with commercially available sensors. In this work, the current gold standards are the chemiluminescence and NDIR methods. The purpose of the Bland-Altman method is to assess the agreement of two methods, where one method is the clinical gold standard and the true value is not known or cannot be easily determined. The Bland-Altman method does not compare the accuracy of the two methods. Instead, the Bland-Altman method is used to determine if the new method has readings sufficiently similar to the standard technique to allow the new technique to replace the standard technique in clinical decision making.

For the offline $\mathrm{NO}$ concentration measurements, a narrower limit of precision $(\pm 2.9\mathrm{ppbv})$ versus the entire dataset $(\pm 4\mathrm{ppb})$ was obtained by dividing the offline $\mathrm{NO}$ samples into two groups according to the order of analysis: 1. ICOS first and chemiluminescence second, and 2. vice versa. To verify the 2- $\sigma$ limits of precision determined using the Bland Altman method, the standard deviations of both instruments were determined at the time of measurement, which yielded $\sigma =1.3\mathrm{ppbv}$ for the chemiluminescent analyzer and $\sigma =1.5\mathrm{ppbv}$ for the ICOS analyzer. The two standard deviations were added quadratically, yielding $2\mathrm{ppbv}$ , which is in reasonably good agreement with the 1- $\sigma$ limits of precision $\left(1.5\mathrm{ppbv}\right)$ estimated using the Bland-Altman method. The standard deviations of the chemiluminescent and ICOS analyzers were determined using the 76-ppb $\mathrm{NO}:{\mathrm{N}}_2$ calibration gas. The ICOS analyzer was undergoing continuous development during the reported studies, and therefore our best to date minimum detectable $\mathrm{NO}$ concentration $\left(0.4\mathrm{ppbv}\right)$ was better than the minimum detectable $\mathrm{NO}$ concentration measured at the time of intercomparison investigations for the offline measurements.

The limit of precision for each group individually had remarkably similar limits of precision $(\pm \sim 2.9\mathrm{ppbv})$ , suggesting that a different systematic error was introduced into each group of data. The systematic error in the offline comparison appeared to be due to a decreasing $\mathrm{NO}$ concentration over the 4 to $6\mathrm{h}$ between consecutive concentration measurements. In the group measured by ICOS first [Fig. 1a], the average difference (ICOS minus chemiluminescence) was greater than 0 $\left(1.1\mathrm{ppbv}\mathrm{NO}\right)$ , whereas in the group measured by chemiluminescence first [Fig. 1b], the average difference (ICOS minus chemiluminescence) was less than 0 ( $-0.7$ -ppbv $\mathrm{NO}$ ). This finding suggests that $\mathrm{NO}$ is chemically unstable and decreases in Tedlar bags during a 4- to 6-h period.

The ICOS offline ${\mathrm{CO}}_2$ measurements were in good agreement with the NDIR capnograph. The average difference was $0.01\%{\mathrm{CO}}_2$ and the limits of agreement were $+∕-0.13\%{\mathrm{CO}}_2$ . The percent ${\mathrm{CO}}_2$ differences are not large enough to be clinically important, and ICOS ${\mathrm{CO}}_2$ measurements can be used clinically.

For online analysis, the $\mathrm{NO}$ plateau concentrations were estimated for breath measurements with both the ICOS and chemiluminescent analyzers over a range of exhalation flow rates (data not shown). The average of $15\mathrm{NO}$ concentration profiles at a $3\mathrm{l}∕\min$ exhalation rate are shown in Fig. 3. The 3-l/min exhalation flow rate was chosen, as it is recommended by the lower airway $\mathrm{NO}$ guidelines.⁶ The $\mathrm{NO}$ plateau regions for ICOS and chemiluminescence were in good agreement, with all data points within 1 standard deviation of the 15 measurements. The ICOS measurements had a larger standard deviation for each averaged concentration.

Tidal breath $\mathrm{NO}$ measurements require subsecond time resolution $(<0.2\sec )$ with a detection sensitivity of $\sim 1\mathrm{ppbv}$ . From Fig. 5, it is evident that the chemiluminescent sensors had adequate sensitivity to detect the plateau $\mathrm{NO}$ from the waveform. For tidal $\mathrm{NO}$ measurements, better detection sensitivity of the ICOS sensor is required to improve clinical decision making.

The ICOS sensor was found to be in good agreement with the NDIR capnograph for online end-tidal ${\mathrm{CO}}_2$ measurements. End-tidal ${\mathrm{CO}}_2$ varies with ventilation, where increased ventilation produces a lower end-tidal ${\mathrm{CO}}_2$ . Variation in ventilation was minimized in this study by using paced breathing during online and tidal ${\mathrm{CO}}_2$ measurements. For online ${\mathrm{CO}}_2$ measurements, the standard deviation of end-tidal ${\mathrm{CO}}_2$ values was similar $(\sigma =0.2\%{\mathrm{CO}}_2)$ for the ICOS and NDIR sensors.

The ICOS online ${\mathrm{CO}}_2$ waveforms began $\sim 2\sec$ after the NDIR waveforms (see Fig. 4). The delay was due to the difference in the position of the sensor in the breath collection system. The NDIR sensor was positioned near the mouth in-line with the exhalation flow, whereas the ICOS sensor drew air from the exhalation flow into the sensor at a rate of $700\mathrm{ml}∕\min$ . The time lag is estimated to be $\sim 2.4\sec$ based on the $700\mathrm{ml}∕\min$ sample flow rate, an $\sim 7\text{-}\mathrm{ml}$ total gas cell volume (at $40\mathrm{torr}$ ), and a 1.8-sec integration time. A time lag was not observed between the chemiluminescent and ICOS sensors because the chemiluminescent sensor also drew air from the exhalation flow.

As with online ${\mathrm{CO}}_2$ analysis, the different positions of the NDIR and ICOS sensors accounted for the faster response time of the NDIR sensor. The NDIR sensor was able to rapidly detected changes in ${\mathrm{CO}}_2$ during tidal breathing, allowing visualization of the three phases of exhaled ${\mathrm{CO}}_2$ , labeled in Fig. 6. The gas handling system of the ICOS sensor did not have sufficient time response to resolve the three phases of expiratory ${\mathrm{CO}}_2$ .⁴¹ The maximum sample flow rate of $0.8\mathrm{l}∕\min$ into the ICOS sample cell restricted the system gas exchange capability and was the limiting factor in the time resolution of the ${\mathrm{CO}}_2$ measurement. At the flow rate of $0.8\mathrm{l}∕\min$ , the time required for a complete gas exchange within the sampling cell is $2.5\sec$ . The volume of the ICOS sensor system includes the sample cell and tubing connecting the breath collection system, the sample cell, and the pump. A higher gas flow and smaller total volume would improve the time response of the ICOS sensor. In the case of breath measurement, in which the expected ${\mathrm{CO}}_2$ concentrations are at levels between 2 and 5%, it is important that the absorption lines are strong enough to have high signal-to-noise ratio but at the same time are not saturated. In this work we were able to target optimal ${\mathrm{CO}}_2$ absorption lines in the $1915\text{-}{\mathrm{cm}}^{-1}$ region that allowed for rapid and sensitive exhaled ${\mathrm{CO}}_2$ measurements.

Simultaneous $\mathrm{NO}$ and ${\mathrm{CO}}_2$ measurements can be made with the ICOS sensor by measuring the $\mathrm{NO}$ absorption line at $1915{\mathrm{cm}}^{-1}$ and the ${\mathrm{CO}}_2$ absorption line at $1915.57{\mathrm{cm}}^{-1}$ , accessible within a single frequency scan of the QCL. The linewidth of the ${\mathrm{CO}}_2$ absorption feature is sufficiently low to prevent spectral interference with the neighboring $\mathrm{NO}$ line. No spectral interferences from other molecules are present in this wavelength region.

Several improvements can be made to the ICOS sensor to achieve better sensitivity of exhaled $\mathrm{NO}$ during tidal breathing and to improve time resolution of ${\mathrm{CO}}_2$ measurement during tidal breathing. Mirrors with higher reflectivity are available and would improve the sensitivity of both gases and allow a smaller ICOS gas sample cell to be used. Quantum cascade lasers are available at $1900{\mathrm{cm}}^{-1}$ , where the strongest, interference-free $\mathrm{NO}$ line in the R-branch of the fundamental ro-vibrational band is located. The $1900\text{-}{\mathrm{cm}}^{-1}$ line is $\sim 1.4$ times stronger than the $1915{\mathrm{cm}}^{-1}$ line used in this work. Wavelength modulation spectroscopy, where the QCL current is modulated at a high frequency and a lock-in amplifier is used to detect the second harmonic frequency, can also be used to improve the signal-to-noise ratio.⁴² Off-axis ICOS sensors have benefited by applying this technique.^{32, 33, 43} Furthermore, a concentration determination method applied in this work, based on a fitting of the measurement data with the tabulated reference spectrum, can benefit from faster data acquisition by increasing the number of points per laser frequency scan, and thereby reducing the fitting uncertainty.

6.

Summary

In this work an exhaled $\mathrm{NO}∕{\mathrm{CO}}_2$ sensor for breath analysis using mid-infrared QCL-based integrated cavity output spectroscopy is compared to a commercial chemiluminescence $\mathrm{NO}$ sensor and nondispersive infrared absorption ${\mathrm{CO}}_2$ sensor. The ICOS sensor shows good agreement with the two commercial sensors using offline, online, and tidal sampling techniques. The ICOS sensor has a noise-equivalent sensitivity $\left(1\sigma \right)$ for $\mathrm{NO}$ of $0.4\mathrm{ppbv}$ with a 1-sec averaging time. Potential improvements to the ICOS sensor include incorporating higher reflectivity mirrors and utilizing the $\mathrm{NO}$ absorption line at $1900{\mathrm{cm}}^{-1}$ . A key advantage of the ICOS-QCL sensor is its ability to simultaneously detect multiple clinically relevant exhaled molecules, in this case $\mathrm{NO}$ and ${\mathrm{CO}}_2$ .