2.1.

Tissue Optics

Tissue, being mainly composed of water, blood, fat, proteins, and skeletal structures, exhibits the tissue optical window ranging from wavelengths of $\sim 600\mathrm{nm}$ to 1.3 μm, in which the light absorption is relatively weak so that light can be reasonably well transmitted although the light scattering is still strong.⁹ The lower wavelength limit is due to the absorption of hemoglobin and oxyhemoglobin,¹⁰ while the upper one is mainly due to (liquid) water absorption.¹¹ In this window, the tissue has optical properties so that, on average, the light traveling in the tissue will change direction completely $\sim 10$ times per centimeter (reduced scattering coefficient, ${\mu }_s^{\prime }\approx 10{\mathrm{cm}}^{-1}$), while $\sim 10$ to 40% of the light is absorbed after every centimeter of travel (absorption coefficient, ${\mu }_a\approx 0.1$ to $0.4{\mathrm{cm}}^{-1}$). When illuminating tissue in this spectral region, it is thus possible to detect diffuse light some distance away on the same surface or on the other side of an object. The path-length that the light has traveled between the position of the source and detector is greatly prolonged due to the scattering.

A major challenge in the field of tissue optics is to understand the light propagation on a detailed level, following light injected into the tissue. Furthermore, it is demanding to calculate the photon-hitting density, i.e., the spatial distribution of the conditional occupancy for light injected in the tissue at one position and detected at another. One way to obtain some information on this, when analytical approaches are difficult or impossible to apply, is to perform Monte Carlo simulations. This type of simulation is used for describing the light propagation in the present paper.

2.2.

Principle of the Spectroscopic Technique

Gas present in pores or cavities embedded in the tissue will create sharp and species-unique absorption imprints in the transmitted light, which are readily observed in the detected signal. The measurement technique GASMAS (Ref. 7) is utilizing these gas absorption features and has earlier been shown to perform well for gas assessment in human sinus cavities for diagnosis of sinusitis^12–15 and in the mastoid bone,¹⁶ as well as in many nonbiological light scattering objects, e.g., wood,¹⁷ pharmaceutical tablets,^18,19 ceramics,²⁰ food packages,^21,22 fruits,^23,24 etc. In short, the technique can be used to evaluate the gas content of small pores or larger cavities embedded in strongly scattering (turbid) media. In this study, the gases of interest are oxygen (${\mathrm{O}}_2$) and water vapor (${\mathrm{H}}_2\mathrm{O}$), which have a number of absorption lines at light wavelengths around 760 and 935 nm, respectively.

In order to probe a gas, a tunable laser is scanned across a wavelength interval where the gas of interest has an absorption line and the light is sent through a path with the gas. A detector records the intensity of the light that has passed though the gas as a function of time. At the time of the scan when the laser wavelength exactly matches the wavelength of the absorption line, a slightly lower intensity is recorded [see Fig. 1(a)]. The fraction of light lost at the absorption line, $A$, is dependent on the concentration, $c$, of the gas and the length of the path that the light goes through the gas, $L$ (the path-length). This relationship is given by the Beer-Lambert law.

Fig. 1

Principle of tunable diode laser absorption spectroscopy and wavelength modulation spectroscopy (WMS).

Here, $I$ is the intensity of light recorded at the absorption line, $I_0$ is the intensity at wavelengths adjacent to the line, and $\epsilon$ is the molar extinction coefficient. Thus, by observing the magnitude of the absorption dip caused by the absorption of the gas, the product of concentration and path-length can be found (the extinction coefficient is constant and well known). To extract the concentration or path-length from their product, the other factor must be known. This is further discussed in Sec. 3.2.

The performance of the detection can be improved by employing the wavelength modulation spectroscopy (WMS) technique. Basically, the wavelength of the laser is quickly modulated at a frequency $f_m$ at the same time as it is scanned across the absorption line. By doing this, it is possible to use powerful lock-in detection approaches. A lock-in amplifier is tuned to the frequency of the modulation, as well as to overtones of this [$1\times f_m,2\times f_m,3\times f_m,\dots$, see Fig. 1(b)]. WMS was employed in this study in a digital manner described thoroughly in Refs. 14 and 25.

3.1.

Subjects of the Case Study

The present work is a pilot study for a project with ethical approval (no. 213–356) by the Regional Human Ethics Committee at Lund University, Sweden. Measurements were performed on three healthy, full-term infants with parental informed consent and presence at the measurements. Relevant data for the subjects are presented in Table 1.

Table 1

Subjects of the study.

3.2.

Technical System

The system used for the spectroscopic study is presented in Fig. 2. Driving voltages consisting of a triangular ramp (scan) with a frequency of 5 Hz, and a superpositioned fast modulation at 9015 and 10,295 Hz for water vapor and oxygen, respectively, are generated by a LabView interface in a standard PC and sent via a data acquisition (DAQ) card (NI-6120 with connector block BNC-2110) to the laser current and temperature controllers (Thorlabs, New Jersey LCD201C and Thorlabs TED200C) controlling the drive current and temperature of the laser for water vapor analysis, respectively, and to a Melles Griot 06DLD103 unit, controlling both current and temperature of the laser for oxygen analysis. One of the lasers (Nanoplus, Gerbrunn, Germany DFB 213/6-24) emits light with a wavelength of $\sim 760\mathrm{nm}$, suitable to probe oxygen, while the other laser (Nanoplus, DFB 111/1-2) emits light with wavelength of $\sim 935\mathrm{nm}$, suitable to probe water vapor. The output power is $\sim 4\mathrm{mW}$ for each of the lasers. Both lasers are fiber coupled and the light emissions from the two lasers are then merged together into a single fiber, thus containing light of both wavelengths. The light is then again split up in two fibers, one containing $\sim 90\%$ of the light (both wavelengths) and the other containing the remaining 10% (both wavelengths). The distal end of the fiber containing most of the light is placed on the skin of the infant and the light is delivered to the desired body area, while the smaller light fraction is sent directly to a reference detector used for signal improvement. The weak light detected by the two photodiodes (Hamamatsu, Hamamatsu, Japan S3204-08, active size $18\times 18{\mathrm{mm}}^2$, and Hamamatsu S3590-01, active size $10\times 10{\mathrm{mm}}^2$, for the in vivo and reference detectors, respectively) will generate small currents that are amplified and transformed into voltages with transimpedance amplifiers (Femto, Berlin, Germany DLPCA-200). The resulting signal voltages are then fed to the same DAQ card as was originally used to generate the inputs to the laser drivers. The voltage signals are recorded with a sampling rate of 400 kHz. The signals from each 0.2-s ramp can be averaged for an arbitrary number of times before being stored for further analysis. The lock-in procedure is digitally performed with a custom-made MATLAB® software, and the final result is a measure of the product of the path-length, $L$, and the gas concentration, $c$, mentioned in Sec. 2.2. The individual absorption signals by water vapor and oxygen are recovered by performing the digital lock-in amplification at the modulation frequencies for the two lasers, respectively. A reference recording through 1 m of air ($\text{temperature}=22°\mathrm{C}$, relative $\text{humidity}=32\%$) was performed in connection with the measurements and was used as a calibration to give quantitative values of the measured gas absorption. The strength of this absorption signal is recalculated as if it would have been performed in 100% relative humidity and at 37°C (the in situ conditions) before being compared with the obtained in situ gas absorption signals. The absolute strength of the recorded absorption signals is then expressed in terms of equivalent mean path-length, $L_{\mathrm{eq}}$, measured in units of millimeter. If the strength of the obtained absorption signal is 10% of that of a recording through 1 m of air with 100% relative humidity and a temperature of 37°C, the result is $L_{\mathrm{eq}}=100\mathrm{mm}$. This way of expressing the result directly gives the (average) physical length that the water vapor probing light passed through gas in the body (assuming the above conditions). Assuming the light probing oxygen has the same path-length through gas¹² enables estimation of its concentration through its absorption magnitude [the only unknown remaining in Eq. (1) is $c$]. Due to the demanding measurement conditions with very low light transmission and small absorption imprint, the signal quality is often relatively poor. The accuracy of the final estimated gas absorption is improved by using the above calibration including a fitting procedure.

Fig. 2

Setup for the measurements. A PC with the software LabView is, through a data acquisition (DAQ) card with a connector box, sending the voltage ramp/modulation inputs to the current drivers. The current drivers are in turn sending the currents to the lasers, which emit the light into optical fibers. The light from the two lasers, being of two different wavelengths, is merged through a multiplexer (MP) and again split into two fibers (both now containing the two wavelengths). One of the fibers goes to a reference detector and one goes to the infant. The probe light is detected a small distance away from the fiber tip. The detector (Det.) converts the light to a current that is amplified and transformed to a voltage in the transimpedance amplifiers (TIA) before being sent to the connector box of the DAQ card to be sampled. The sampled voltage signals are read by the computer and stored for analysis.

4.1.

Ultrasound Measurements

Ultrasound measurements (SonoSite, Washington M-Turbo^®, Linear probe, 6 to 13 MHz) were conducted on the infants to obtain images of the distance between the skin surface and the organs at interest for investigation, i.e., the thickness of the tissue surrounding the lungs and intestine, as well as the depth of possible gas pockets in the intestine. An experienced ultrasonographist performed the assessments.

The measurements on the lungs were made with the ultrasound probe placed in different positions, pointing toward the lung tissue. However, most measurements were conducted with the probe placed slightly below the collar bone [similar to the placement of the fiber in Fig. 3(a)].

Fig. 3

(a) Geometry of the lung measurements that gave gas imprints. (b) Front geometry of the intestine measurements. (c) Back geometry of the intestine measurements. The laser light was delivered through the thin fiber and detected with the photodiode connected with a BNC cable.

For the abdominal cavity, several measurements were obtained with the ultrasound probe placed on the front side of the stomach [similar to the placement of the fiber and detector shown in Fig. 3(b)].

4.2.

Spectroscopic Lung Measurements

The tissue surrounding the lungs of the three infants was (see Fig. 4) relatively thick and therefore difficult to penetrate with the laser light. The measurements presented in this paper therefore aimed toward trying to find a laser/detector position geometry where it was possible for the light to go deep enough to actually probe the gas in the lung and at the same time be possible to detect at another position. Numerous geometries were tested with the laser/detector at the back, chest, and side of the torso. The tissue covering the lungs is thinnest slightly below the armpit (low fat content and little muscle tissue between the skin and the ribs)—hence, it was with the detector positioned here that it was possible to detect a small absorption signal by water vapor. The geometry that was found to function best for all of the subjects is shown in Fig. 3(a), where the laser light is sent in just below the collar bone and detected through the ribcage, just below the armpit. This geometry was used on the right side of the body, sending light through the right lung.

Fig. 4

Ultrasound image, showing an $\sim 2\text{-}\mathrm{cm}\text{-}\text{deep}$ thoracic wall cross-section of the three infants [(a) to (c), corresponding to subjects 1 to 3]. The probe was placed 2 cm below the right collar bone. As indicated, skin surface, ribs, pleura, and lung tissue are clearly seen.

4.3.

Spectroscopic Intestine Measurements

Several different positions for placing the fiber tip and the detector were tested to search for any sign of oxygen or water vapor. Oxygen was not expected to be present, but if any gas-filled volumes were forthcoming, water vapor should definitely be a part of the gas. When one such source-detector configuration was found in the case of subject 1, a special study was performed where the strength of the absorption signal by water vapor was evaluated at four different fiber-to-detector separations: 20, 31, 43, and 54 mm (fiber tip to center of detector). In this geometry, shown in Fig. 3(b), the fiber probe (delivering the laser light) was positioned roughly at the height of the umbilicus and the detector was placed on the abdomen to the side of the fiber probe (at the four different separation distances). For subject 2, measurements were performed in the geometries shown in Figs. 3(b) and 3(c), while only geometry “b” was used for subject 3.

4.4.

Simulations

Simulations were used to investigate the light transport in relation to the measurements where the fiber-detector distance was increased stepwise for the intestine measurements on subject 1. The simulations utilized a custom-made Monte Carlo program injecting single photons (i.e., not weighted photon packages) into a two-dimensional grid consisting of $200\times 400\text{blocks}$ ($\text{depth}\times \text{horizontal direction}$). The grid was 4 cm in depth and 10 cm in the horizontal direction. The length of each photon step was randomized in a logarithmic distribution with center-of-mass length $\langle l\rangle =1/{\mu }_s^{\prime }=0.1\mathrm{cm}$, and the direction was completely random in 360 deg. For each step, the probability of absorption was ${\mu }_a\times l$, with ${\mu }_a=2.5{\mathrm{cm}}^{-1}$. When 200,000 photons had reached the detector, the simulation was stopped and the paths of these photons were stored (the number of photons injected depends on the position of the detector, but, in general, hundreds of millions of photons were needed). General references for the simulation technique in the tissue optics field can be found in Refs. 26, 27, and 28.

5.1.

Ultrasound Results

Example results from the ultrasound measurements toward the lungs of the subjects are shown in Fig. 4. The measurements performed should indicate the thickness of the tissue that the light has to penetrate in order to reach gas. Although it is quite difficult to get exact measures of the depth since the placement of the probe is critical, the results indicated that the lung tissue (not necessarily gas), in this geometry, started at depths of 1.15, 1.53, and 1.03 cm for subjects 1, 2, and 3, respectively.

Gas-filled regions were found in the intestine in some of the ultrasound investigations performed on subjects 1 and 2, while no such clear findings could be seen for subject 3. Figure 5 shows one example of an ultrasound image of the intestine on each of the three subjects, down to a depth of $\sim 2\mathrm{cm}$. As seen, the gas pockets found for subjects 1 and 2 are located at depths of $\sim 1$ to 2 cm.

Fig. 5

Ultrasound images showing an abdominal cross-section of the three infants [(a) to (c), corresponding to subjects 1 to 3], $\sim 2\mathrm{cm}$ of depth. The ultrasound probe was placed adjacent to the umbilicus. Skin surface, skeletal muscle, and air-filled intestine are indicated in the figures. In one of the images (c), due to normal bowel movements, a cross-section of air-filled intestine was not possible to capture.

5.2.

Lung Spectroscopic Results

For most laser/detector geometries, the light intensity received by the detector was extremely small due to the thick tissue. Further, in most positions, where the light transmission was reasonable, no absorption signal could be seen, since the light did not probe sufficiently deep to reach the gas in the lungs. However, in one position, as described in Fig. 3(a), it was possible to obtain WMS absorption signals by water vapor that was clear enough to distinguish with certainty for all subjects. The $2f$-demodulations of these absorption signals are presented in Fig. 6.

Fig. 6

Second harmonic demodulation of signal received with the light sent in just below the collar bone and detected below the armpit, as illustrated in Fig. 3(a). The figure shows the in situ signals (noisy) with fitted reference recordings (no noise). The SNR is low but enough to ensure contact with gas. Signal (a) is obtained from a measurement on subject 1, with an integration time of 60 s, (b) is from a measurement on subject 2, with an integration time of 40 s, and (c) is from subject 3, with an integration time of 60 s. The SNR is about the same in all recordings, even though the resulting $L_{\mathrm{eq}}$ values are very different, which can be ascribed to the fact that the received light intensity was much lower in the case of subjects 1 and 3, than in subject 2. The much smaller absorption signal received for subject 2 could still reach the same SNR as that of the others.

5.3.

Intestine Spectroscopic Results

A variety of configurations in the positioning of the fiber tip and detector were tested in searching for gas imprints from the intestines. We found that, for subject 1, most positions did not give gas signals, neither for oxygen nor for water vapor. However, a couple of geometries were found where clear signals by water vapor were seen. Gas-filled cavities in the intestine had therefore been found. As a special case study, the signal strength was evaluated at four different fiber-to-detector separations. The results are presented in Fig. 7. The equivalent path-lengths presented indicate the average physical length that the light has passed through gas in the body.

Fig. 7

Water vapor absorption (second harmonic WMS) signals and average received intensity obtained on the abdomen (geometry “b”) with four fiber-to-detector separations. It is clearly shown how the absorption signal is increasing with increased separation, while the intensity is decreasing. The origin to these effects is shown in Fig. 9. The curves were each obtained from data obtained during 30 s.

For subject 2, water vapor absorption signals were found, both in the same geometry as for subject 1, i.e., with the fiber tip and detector placed on the stomach [Fig. 3(b), geometry b], and also with the fiber and detector placed on the lower part of the back [Fig. 3(c), geometry c]. For subject 3, intestinal gas signals were only received in geometry b. Figure 8 presents example results obtained in geometries b and c for subject 2, and in geometry b for subject 3.

Fig. 8

WMS signals obtained on the intestine of subjects 2 and 3. The data in (a) is recorded on subject 2 in geometry “b” according to Fig. 3, (b) is recorded in geometry “c” on the same subject, and (c) is obtained from a recording in geometry “b” on subject 3. All recordings were done with an effective integration time of 10 s. Again, the SNR is not only related to the $L_{\mathrm{eq}}$-value, but also to the received intensity, etc. As previously mentioned, the traces without noise show the fitted reference recordings.

As seen in Fig. 7, the acquired absorption signals by water vapor were strong and easily distinguished. The dependence on fiber-detector separation for the light intensity and absorption signal strength is further illustrated in Fig. 9. The origin of this effect is partly due to the fact that the light must travel a longer distance through tissue and probably also gas to reach the detector, and partly because a larger separation increases the probed depth into the tissue. The increased depth provides an increased probability of light reaching into the intestine. This separation-dependent penetration depth is shown with the help of results obtained by Monte Carlo simulations together with the same experimental data as shown in Fig. 7. The figure shows the distribution of the conditional light injected through the fiber and detected by the detector. (Note that light that is emitted by the fiber tip is scattered in all directions with equal probability, while the presented profiles are only considering light that actually reaches the detector.) The parameters used in the simulations were ${\mu }_s^{\text{'}}=10{\mathrm{cm}}^{-1}$ and ${\mu }_a=0.25{\mathrm{cm}}^{-1}$.

Fig. 9

Photon hitting density results from a Monte Carlo simulation of light injected into the tissue at one point and detected a certain distance away with an extended detector area. The profiles shown indicate the probability distribution of where the light that reached the detector went, on its way, from fiber to detector. As the distance between the detector and fiber is increased, the light probes deeper into the tissue (the detected light intensity is also rapidly decreasing). The density profiles shown to the right indicate the depth profile of the light penetration at the center plane between the fiber and detector and the cross indicates the average light depth at that plane. The gas absorption signals, or WMS signals, are thus increased with fiber-to-detector separation. Projections of the WMS signals are provided in the back plane for clearer intercomparison.