1. Introduction

Respiratory function is one of the most important health metrics in healthy individuals. Normal human respiratory gas is a mixture of various gases containing hundreds of volatile organic compounds (VOCs) and large amounts of nitrogen (N₂), water (H₂O) and carbon dioxide (CO₂). These gases derived from the blood by passive diffusion across the pulmonary alveolar membrane, and variation in the concentration of biomarker molecules is closely related to various metabolic diseases [1–4]. Breath gas analysis is intrinsically safe and noninvasive, opens a direct and unique window onto the blood composition, and offers a new approach for screening and early diagnosis of respiratory disease. Increasing interest has been expressed in the possibility of breath analysis as diagnostic and monitoring tools [5,6]. For instance, exhaled nitric oxide (NO) was demonstrated to be an indicator of asthma or airway inflammation [7,8], acetone was found to be more abundant in the breath of diabetics [9–12], exhaled ammonia level is significantly elevated in patients with renal diseases [13], and increased alkanes and formaldehyde concentration is present in the case of lung cancer and breast cancer [14]. For patients with asthma, chronic obstructive pulmonary disease (COPD), pneumonia or sleep apnea, the lungs cannot remove enough of the CO₂ produced by the body, and respiratory acidosis may occurs [15]. A measurement of CO₂ concentration is a primary approach to monitor the amount of lung ventilation. Moreover, monitoring the exhaled CO₂ can reduce the potential risks for sudden infant death syndrome (SIDS) and it is also useful for primary assessment of the respiratory health of critically ill patients [16]. Furthermore, recent investigations have shown that reduced end-tidal CO₂ (EtCO₂) pressure is a strong predictor of adverse events in cardiac disease [17].

In all these applications, a fast, accurate and precise measurement of exhaled breath component is essential. A variety of measurement techniques for trace molecular detection have been studied extensively, including laser absorption spectroscopy [18–21], mass spectrometry and electronic noses [22]. Laser spectroscopy offers unique advantages in terms of sensitivity and specificity, as well as fast response. Zare et al. studied the isotopic ratio of ¹³CO₂ to ¹²CO₂ in the wavelength near 1.6 µm by the cavity ring-down spectroscopy (CRDS) [23]. Zhen Wang et al. developed quartz-enhanced photoacoustic spectroscopy (QEPAS)-based isotopic sensor using mid-infrared interband cascade laser (ICL) [24]. However, the CRDS needs long gas chambers and high-quality cavities, it’s hard to develop compact and low-cost miniaturized system; the microresonator of QEPAS spectrophone must be enclosed, restrict its practical application for online measurement. Tunable diode laser absorption spectroscopy (TDLAS) is also a versatile technique for unambiguous measurements of trace gas due to its narrow spectral resolutions [25]. Single mode laser operation along with a broad tuning range is highly desirable for TDLAS. In general, the CO₂ has weaker absorption intensity in near-infrared wavelength region than that in mid-infrared, while most optical devices in the near-infrared range are simple and can operate within room temperature, it is very attractive for the development of reliable, compact and low-cost measurement system. Vertical-cavity surface-emitting laser (VCSEL) has the merit of low power consumption, fast frequency responses and less expensive, and it has a wider wavelength tuning range than DFB laser, provides more choices to select the suitable absorption lines for gas sensing [26,27].

In the present study, a tunable VCSEL diode emitting around 1579 nm is used for the detection of exhaled CO₂. In combination with a 20 m path-length Herriott-type cell and 2f wavelength modulation spectroscopy, a minimum detectable concentration of 0.769‰ was obtained. Different CO₂ concentrations were obtained in dead space and alveolar gas of healthy subjects, and the exhaled CO₂ increased significantly with increasing physical activity, reached its maximal value at the beginning of respiratory compensation, and then decreased slightly until maximal exercise.

2. Line selection

Human exhaled breath consists of 79.5% N₂, 14.2% oxygen (O₂) and 4% CO₂ [28,29], a specific wavelength should be selected to be free from interference of those large-abundance molecules. Moreover, trace gases such as CH₄, SO₂ and CO can be found in the exhaled breath [30–32], but usually only in very small amounts. In addition, the exhaled air has a relative high humidity, which might introduce measurement errors, the absorption bands of water molecules should be avoid for highly sensitive detection. According to the HITRAN molecular spectroscopic absorption database [33], the CO₂ molecule shows several strong absorption bands in the near-infrared spectral regions, and the absorption line centered at 1579 nm was selected as the optimum target line [34,35]. Figure 1 depicts the absorption spectra of CO₂, H₂O and O₂ at atmospheric pressure and 296 K. It is evident that the spectral line intensity of CO₂ located at 1579.57 nm (6330.82 cm⁻¹) is determined to be 1.5×10⁻²³ cm⁻¹/(molec·cm⁻²), meanwhile, the line intensities of O₂ and H₂O is 4.5×10⁻³¹ cm⁻¹/(molecule·cm⁻²) and 1.6×10⁻²⁷ cm⁻¹/(molecule·cm⁻²), ∼7 and 4 orders of magnitude lower than that of CO₂, respectively.

 

Fig. 1. The absorption lines of O₂, H₂O and CO₂ near 1579 nm taken from the HITRAN database [33].

Download Full Size | PDF

Table 1 summarizes various exhaled breath gas concentrations and the HITRAN-based absorption line intensity, and the calculated absorbance at 1579.57 nm, temperature of 296 K and 1 atm pressure with an optical path-length of 20 m. As depicted in this table, exhaled breath contains large amounts of N₂, O₂ and H₂O, while both the absorption line intensities and absorbances within 20 m optical path-length are several orders of magnitude lower than that of CO₂. The VOCs gases in exhaled breath are only in very small amounts, their contributions to optical absorption has been found to be negligible. Therefore, the selected laser wavelength can be used for reliable detection of CO₂.

Table 1. The concentration and absorption characteristics of exhaled breath to laser at 1579.57 nm.

View Table

3. Experimental details

The experimental schematic is shown in Fig. 2. A VCSEL diode (VERTILAS, VL-1579-1-SQ-A5) emitting in the 1579 nm waveband is used as the light source. It can scan over a spectral range of 1578.12-1583.41 nm by tuning the temperature and injection current, and the maximum output power is 3 mW. The temperature and current of the laser are adjusted using a compact driver board (Wavelength, LDTC02020), all the controls and indicators are onboard for simple plug-and-play operation. The laser was scanned across the absorption lines using sawtooth waveform supplied by a second harmonic (2f) modulation board. In order to perform wavelength modulation spectroscopy (WMS), a sinusoidal waveform generated by the 2f modulation board was superimposed on the sawtooth waveform. The output laser from VCSEL diode is connected to a collimator (Thorlabs, A375TM-C), passing through a wedge window and then injected into the Herriott cell (Lambert Technology, Shenzhen, China) with 32 cm base length, 20 m effective optical length and capacity of 420 cm³. Finally, the output laser beam from the gas cell is detected by InGaAs photodetector with focused lens (Thorlabs, DET08CL/M). The output signal from the detector was amplified by a custom made transimpedance amplifiers (TLC2201, Texas Instruments) and sent to a lock-in amplifier for 2f demodulation. Finally, the 2f signal is collected by digital oscilloscope (MDO4054C, Tektronix) with a LabVIEW program and analyzed by Origin and Matlab software.

 

Fig. 2. Schematic drawings of the experimental TDLAS setup.

Download Full Size | PDF

The wavelength modulation amplitude was appropriately chosen in the measurements in order to obtain the strongest spectral signal. The sinusoidal wave used for modulation had a peak-to-peak voltage of 5 mV and a frequency of 10 kHz. The peak-to-peak amplitude of the sawtooth wave was 100 mV, which scanned the laser diode current from 6.2 to 7.2 mA (10 mA/V) at a cycle length of 400 ms, offers 0.8 mW average output power. The absorption line intensity is strongly dependent on the temperature, and the absorption line width is pressure dependent, these parameters directly affect the peak value of 2f signal and accuracy of gas concentration measurement. In the experiment, the pressure and temperature in the cell were controlled using a commercial gauge (AP-C30WP, Keyence) and a temperature controller (XH-W1308, Xinghe Electronics), respectively. The visible light from the alignment diode laser (λ=630 nm) was employed as a guide beam to assist the alignment of optical system. To estimate the performance of VCSEL-based TDLAS apparatus for CO₂ detection in exhaled breath, online monitoring of exhaled CO₂ is demonstrated under the optimum pressure. As water vapor attached to the high reflection mirror will affect light transmission, direct exhalations into a mouthpiece are dehumidified by means of a cooling trap at a temperature of 240 K.

4. Experimental results and discussions

To evaluate the performance of the TDLAS system, the absorption line of CO₂ at wavelengths near 1579 nm was measured using a wavelength modulation technique. As N₂ has negligible absorption in the spectral region of interest, pure N₂ (99.99% purity) is first introduced into the gas chamber to remove the original gas in the chamber for each measurement. Then, mixtures of CO₂ (99.99% purity) and N₂ at different concentration levels were introduced into the gas chamber at a pressure of 2.5 kPa. The temperature and pressure were maintained a constant value in each measurement cycle. A typical 64-averaging raw absorption signals at 6330.82 cm⁻¹ were depicted in Fig. 3(a). The voltage of sawtooth wave is proportional to the operating current of the laser diode, which tunes the laser's wavelength linearly. A distinct change of sawtooth waveforms that proportional to the gas concentration was obtained in Fig. 3(a), indicating the significant absorption of CO₂. The 2f signals of CO₂ at 1579.57 nm with concentration of 2.5%, 5.0%, 6.0%, 8.0% and 10% were derived by digital lock-in amplifier technology [36] and illustrated in Fig. 3(b). It is evident from the Fig. 3(b) that as the concentration changes, the intensity of the 2f spectral signal changes correspondingly. The peak value of measured 2f signal was extracted and the fitting curve versus CO₂ concentrations was obtained and displayed in Fig. 3(c). As shown in the figure, the 2f signal intensity has a good linear relationship with the gas concentration in the measurement range, and the linear fitting coefficient R = 0.99829. Because the gas concentration is linear with peak value of 2f signals, the concentration of detected gas can be obtained from the linear equation.

 

Fig. 3. (a) The direct absorption signal of pure N₂ and CO₂ at different concentrations; (b) The measured 2f signal of CO₂ with different concentration; (c) The relationship between the intensities of 2f signal and the concentration of CO₂ at 2.5 kPa, 40-sweep average was used for minimizing the random error of each point.

Download Full Size | PDF

To further investigate the stability and precision of the system, continuous measurement of CO₂ with concentration of 2.5% was performed over a period of ∼1 h. Figure 4(a) presents a typical time series raw 2f signals recorded with no signal processing. As shown in the figure, the 2f peak value was 0.3507, and the mean square of noise was 0.0036, the SNR was calculated to be 97.42, which implied a detection limit (3σ) of 0.769‰. Figure 4(b) shows 10,000 data points concentration levels extracted from 2f signals, the magnitude changes within the range of 2.256% to 2.749%, the mean value was 2.501% and the standard deviation was calculated to be 0.091%. The main sources of drift are believed to be changes in the laser power and line width caused by temperature fluctuations [37]. The Allan deviation as a function of the integration time is also exhibited in Fig. 4(b). The Allan plot gives a precision of 0.91‰ (1σ) without time-domain average, when the integration time increases to 168 s, a minimum value of the Allan deviation is reached, corresponding to a precision of approximately 100 ppm.

 

Fig. 4. (a) 2f signals of CO₂ with concentration of 2.5% and determined at a pressure of 2.5 kPa; (b) The recorded 10,000 data points concentration levels for a period of 4000 s and the corresponding Allan deviation plot.

Download Full Size | PDF

A series of online measurement were carried out to analyze the relationship between CO₂ concentration in exhaled air and human activities condition. As the alveolar breath from deep in the lungs, it reflects concentrations in blood more accurately [38]. In the experiment, a single exhalation breath samples were collected from individual of male participants, 23 years of age. A Gas flow meter enables measurements to be taken at a constant flow rate of 400 cm³/s and the cell pressure maintains a constant of 2.5 kPa by vacuum pump. A typical healthy individual can produce 3500 cm³ in a single exhalation, so that the gas cell volume is flushed constantly in a single online measurement. The recorded 2f signal in Fig. 5(a) begins to rise as the breath is exhaled, as shown in the red line, the amplitude envelope increase from 0.075 to 0.28 and then have a tendency to saturate at end of breath. Some fluctuates abruptly were observed due to the pressure fluctuation of gas cell. The CO₂ exhalation profiles during tidal plus expiratory reserve volume breathing was obtained and displayed in Fig. 5(b). In the figure, phase I shows the initial stage of expiration, phase II shows a sharp rising of waveform slope, phase III represents alveolar plateau [39]. The observed EtCO₂ concentrations at phase III were ∼4 times higher than dead space gas during the exhalation period (Phase I), and then were saturated, the results are well in consistence with the previous studies [40]. Finally, normal breathing in resting state were performed, a typical serious of the recorded 2f peak value within 1 minutes were depicted in Fig. 5(c) without data smoothing. The good cycle repeatability confirms the real-time online detection ability of the system.

 

Fig. 5. (a) Sequences of CO₂ expirograms obtained from healthy 23-year-old male individual during tidal plus expiratory reserve volume breathing, each profile corresponds to a raw 2f signal; (b) Curves of recorded the exhaled CO₂ concentration in three-phase of respiratory rhythm during a single exhalation; (c) Curves of recorded the exhaled CO₂ concentration during normal breath at rest within 1 minutes.

Download Full Size | PDF

The relationship between EtCO₂ concentration and level of physical activity was investigated experimentally. Breath samples were collected from two healthy postgraduate volunteers at rest and during exercise. Figure 6 shows the photograph of exhaled breath measurement in healthy subject during indoor cycling with different workloads. First, a stable baseline was established by introduce pure N₂ into the gas chamber, then the subjects breathe room air at rest and the exhaled CO₂ was real-time monitored. Subsequently, the participants cycled on a bicycle dynamometer (VIAsprit 150p, CareFusion GmbH), increasing the workload by 20 W per minute and finally up to 220 W. Figures 7(a) and (b) show the temporal evolution of the continuously recorded exhaled CO₂ concentration at rest and during exercises. It can be seen from Figs. 7(a) and (b) that as the exercise load increases, the exhaled CO₂ concentration increases gradually, arising from the increased rate of CO₂ production and delivery to the lungs. The peak values of 2f signals of EtCO₂ under different loads were extracted, and the mean value was plotted in Fig. 7(c). It can be seen from the curves that as the load increases, the overall trend of the exhaled CO₂ concentration increased during the early stage and then decreased slightly, suggesting the same trends of EtCO₂ in exhaled breath. Our results show good consistence with previous studies that EtCO₂ reaches its maximal value at the beginning of the respiratory compensation for the lactic acidosis; thereafter, it has a slightly decrease until maximal exercise [41]. The amount of exhaled CO₂ for two volunteers at peak exercise (volunteer 1 at 100 W, volunteer 2 at 140 W) increased 58.58% and 108.61%, respectively, when compared to the resting state. Then the subject will get into fatigue (anaerobic exercise), the exhaled CO₂ content will gradually decline to an individual's equilibrium value, due to the regulation of physiological functions of the human body. Because of different physical fitness and exercise status, there is a difference in the time of entering fatigue, but the overall trend of exercise is roughly similar.

 

Fig. 6. Photograph of the collection and measurement of CO₂ in exhaled breath of healthy subject with workloads.

Download Full Size | PDF

 

Fig. 7. (a) and (b) Trends in the exhaled CO₂ of two volunteers over time during a rest and exercise from 0 to 220 W; (c) Extracting the mean value of the EtCO₂ concentration at different loading levels of two volunteers

Download Full Size | PDF

Gas concentration measurement based on the near-infrared TDLAS has been paid more and more attention, the application of VCSEL diode offers the feasibility of developing reliable, compact, low power consumption and low-cost devices for exhaled breath monitoring. In our experiment, the CO₂ concentration was detected with a VCSEL diode at 1579 nm, and a minimum detectable CO₂ concentration was 0.769‰. As displayed in Fig. 4(a), the time response of the 2f signals was about 10 ms, an order of magnitude lower than that of capnographs (∼130 ms), so the TDLAS system has a high response speed. Overall, the current precision of the VCSEL-based TDLAS system is clearly sufficient to identify changes in concentration of human exhaled CO₂ in real time. Compared with the small-diameter VCSEL diode, the Herriott cell employed in this experiment may be too large, we will further improve the system design for portable health monitoring by employing compact dense-pattern multipass cell and hollow-fiber gas cell in the future. This experimental procedure is of great significance for the detection and separation of alveolar gas in human breath. Similar systems are feasible for the detection of VOCs such as ammonia, methane and acetone by changing optical wavelength. While most of the VOCs are at very low concentrations ranging from part-per-trillion to part-per-billion levels [42], for the quantitative analysis of mixed exhaled gas, dilution-related measurement errors due to dead space admixture should be considered. The CO₂-triggered acquisition may be a feasible strategy for separation VOCs from dead space to get better reflection of the body metabolic function.

5. Conclusions

Human exhaled breath gas sensing has become very effective and highly valuable tool for biomedical diagnosis. Based on TDLAS technique and VCSEL diode, measurements of exhaled CO₂ at the wavenumbers of 6330.82 cm⁻¹ was studied, and a minimum detectable concentration of 0.769‰ for CO₂ detection was obtained. Real-time measurements were carried out for the detection of CO₂ expirograms from healthy subjects at rest and during exercise, different concentrations of CO₂ were obtained in dead space and alveolar gas, and the exhaled CO₂ increased significantly with the increasing physical activity. An increment of 58.58% and 108.61% in EtCO₂ concentration of two subjects was obtained respectively at peak exercise compared to the resting state. The results indicate that the very accurate low-cost compact measurement system has a great potential to be applied in practice for the detection of pulmonary diseases associated with CO₂ retention.