2.1.

Phantom Preparation

The middle ear phantom was constructed with an enclosed tissue cavity formed from a fish swimming bladder, which was used as the “middle ear cavity.” The total volume of the cavity was about 2 mL, which is in close proximity to the volume of a real human tympanic cavity (1 to 2 mL). The tissue-like material (with regard to scattering properties) surrounding the cavity was made from a mixture of water (89%), gelatin (10%), and ${\mathrm{TiO}}_2$ (1%). The detailed fabricating process of the phantom was performed as illustrated in Fig. 2. First, gelatin was added to water in a small glass beaker, which was placed in a larger water-filled beaker for heating [Fig. 2(a)]. Considering that the melting point of gelatin is at about 35°C, the temperature was set at 50°C.³⁰ The solution was stirred until the gelatin was completely dissolved and the solution appeared almost transparent. Then ${\mathrm{TiO}}_2$ was added to the solution which was maintained at a temperature of 50°C [Fig. 2(b)]. Finally, an expanded fresh fish swimming bladder was put into a holder, and the solution was poured over it [Fig. 2(c)]. Partial solidification occurred at room temperature, and the process was repeated until the covering layer was about 3-mm thick [Fig. 2(d)].

Fig. 2

Illustration of the fabricating procedure of the middle ear phantom. (a) The different steps include the heating process for melting the gelatin, (b) adding the TiO2 powder, and (c) successively building up the scattering material around the phantom by pouring the dissolved solution are shown, (d) The ear phantom with a 3-mm thick tissue-like light-scattering layer is shown where a 10-mm diameter uncoated area (bottom) was left to mimic the eardrum.

2.2.

Construction of the Fiber-Optic Probe

The fiber-optic probe used for GASMAS measurements was constructed with nine optical fibers, where one fiber with the core diameter of $600\mu \mathrm{m}$ was used as transmitting fiber and the other fibers with a core diameter of $400\mu \mathrm{m}$ were used as collecting fibers, as displayed in Fig. 3. The fibers were manufactured by Nanjing Chun Hui Science & Technology Industrial Co., Ltd., China. All the fibers were coated with plastic and had a total diameter of 1 mm. Since reflected light without any gas absorption can dilute the signal, the distance between the transmitting fiber and the collecting fibers must be optimized to enhance the fraction of light which has passed through the gas-filled space behind the eardrum, while observing the constraint that the total diameter of the probe must be compatible with the external auditory canal diameter. The considerations follow the procedures discussed for the case of sinus cavity monitoring in backscattering.³¹ After extensive testing, a 0.5-mm spacing distance between the coated transmitting and collecting fibers resulting in a total probe diameter of about 5 mm was selected for optimized measurements (see Fig. 3).

Fig. 3

Schematic drawings of the fiber-optic probe tip, showing (a) a longitudinal view and (b) a transverse view of the probe used for GASMAS measurements of free gas located behind the “eardrum” in a back-scattering geometry.

2.3.

Experimental Setup

A schematic drawing of the GASMAS setup used for gas detection is shown in Fig. 4(a). Two diode lasers (LD-0760-0100 and LD-0937-0100, Toptica, Munich, Germany) operating at the oxygen absorption line at 760.445 nm and the water vapor absorption line at 937.405 nm, respectively, were used as light sources. Laser current and temperature controllers (LCD 201C and TED 200C, Thorlabs, Newton, New Jersey) were used to control the drive current and the temperature of the diode lasers. The controlling current of the laser was superimposed with a 5-Hz ramp wave as well as with a 10,295 Hz sinusoidal wave for oxygen measurement and a 9015 Hz sinusoidal wave for water vapor measurement, which were created by a LabVIEW program. Among these superimposed signals, the ramp wave was utilized to scan the absorption line and the sinusoidal wave was used to modulate the laser to acquire the harmonic signals for software-based lock-in detection. The modulated laser light was injected into the ear phantom through the transmitting fiber and the backscattered light received by the collecting fibers [see Fig. 4(a)] was detected by a photodiode (S3204-08, Hamamatsu Photonics, Hamamatsu, Shizuoka, Japan). The current signal from the photodiode was converted to a voltage signal and then amplified by a variable gain low-noise current amplifier (DLPCA-200, Femto, Berlin, Germany) before being recorded by the input channel of the data acquisition (DAQ) card (NI6120, National Instruments, Austin, Texas).

Fig. 4

Schematic drawing of the two set-ups used for the measurements on the ear phantom. (a) GASMAS measurements in a backscattering geometry of gas behind the “eardrum” and (b) diffuse reflectance measurements on the “eardrum.”

The experimental setup for reflectance spectral measurements is illustrated in Fig. 4(b). Broadband light from a tungsten lamp was transmitted to the ear phantom through a $400\text{-}\mu \mathrm{m}$ core diameter fiber. The reflected light from the surface of the “eardrum” was captured with a reflectance probe (Ocean Optics, Dunedin, Florida) and then detected by a portable spectrometer (USB4000, Ocean Optics). The obtained data were analyzed by the Ocean Optics SpectraSuite software and transformed to the spectral profile. It can be seen from Fig. 4(b) that the reflectance probe in its present implementation is different from the fiber-optic probe used for GASMAS measurements. The reflectance probe was constructed with $13\times 400\mu \mathrm{m}$ fibers where one was used as transmitting fiber and the rest of the fibers were used as collecting fibers, and all the fibers were integrated in a compact way. Before the measurements, the system was calibrated for intensity using a calibration lamp (IES 1000, Labsphere, North Sutton, New Hampshire) and for wavelength with a low-pressure mercury lamp.

2.4.

Data Analysis

The recorded GASMAS data were analyzed with digital lock-in techniques to get the second harmonic ($2f$) signal. Then the $2f$ signal was normalized by dividing it with a direct signal (the signal without gas absorption) to achieve the normalized $2f$ signal, which is insensitive to the amount of light detected by a photodiode. A detailed discussion of the digital lock-in detection and normalization is given in Ref. 32. Considering the influence of optical interference fringes which come from optical elements in the light path, a small vibration motor was used to mechanically modulate and thereby shift the positions of possible etalon fringes to minimize spurious interference.³³ Minute remaining modulations in the detected light could otherwise mimic absorption lines with similar line width. In this study, a matched version of the reference signal as measured by sending the laser directly to the detector was subtracted from the sample signal obtained from the beam which passed the sample to yield a so-called balanced detection signal.³⁴ The matching intensity is found by fitting the reference signal to the sample signal at frequency regions outside the expected oxygen imprint. A sloping signal background could in this way be accurately removed.

Due to the strong scattering of the light travelling in tissue, the path length is unknown, and the gas concentration is not easy to be estimated by the Beer–Lambert law. The obtained $2f$ signal was fitted to a reference signal which was measured over an ambient atmosphere distance, and a so-called equivalent mean path length, $L_{\mathrm{eq}}$, could be obtained, corresponding to the path length that the light would travel in ambient air to get the same oxygen or water vapor absorption imprint. The $L_{\mathrm{eq}}$ can be described by the equation

3.1.

GASMAS Measurements

Before performing the measurements, the ear phantom was perforated to obtain a small orifice and then left in ambient conditions for 24 h. Thus, we mimic the condition of a tympanic cavity in which the gas composition is the same as that of the atmospheric air as regulated by the Eustachian tube. Measurements of oxygen and water vapor were performed sequentially with the probe positioned against the small part of the fish-bladder wall, which was not covered with scattering material (our “eardrum”). Molecular oxygen was measured first, followed directly by water vapor measurement with the same procedure. All the measurements were performed in ambient room conditions with a temperature of about 23°C and a relative humidity of about 60%. During the measurements, each recorded signal was averaged for about 1 min. A typical recorded $2f$ oxygen signal from the measurement on the ear phantom and the fitted curve from the measurement on a 1100 mm path of ambient air are shown in Fig. 5(a). A water vapor signal from the ear phantom measurement and the fitted curve from ambient air are shown in Fig. 5(b). These signals correspond to an absorption fraction of about $1.8\times {10}^{-4}$ and $1.3\times {10}^{-3}$ for oxygen and water vapor, respectively.

Fig. 5

(a) Typical measured $2f$ signal of oxygen and (b) water vapor obtained on the ear phantom using backscattering detection. The signals were integrated over 60 s.

Interpreted from these results, GASMAS shows a good performance using the special probe in a backscattering detection geometry, even though the measured oxygen signal is intrinsically weak. Studies of gas exchange in the phantom were performed for assuring that the recorded GASMAS signals were in fact due to the gas behind the membrane. Now, the cavity was flushed with pure oxygen or nitrogen through the small orifice until the cavity was enriched with oxygen or nitrogen, and then the orifice was sealed. Due to the liquid water in the surrounding tissue, a humidity of about 100% and a temperature close to 37°C is expected in the real human tympanic cavity. Therefore, some liquid water was injected into the phantom cavity to achieve 100% humidity in the middle ear phantom cavity before the water vapor measurements. Continued measurements were then performed for each gas, integrating the signal and reading out the result intermittently, to create a recording of gas content as a function of time. The orifice was opened after measurements for 5 min (as indicated in the figures) in order to study subsequent gas exchange. The experimental results are shown in Fig. 6.

Fig. 6

Studies for gas exchange after perforation of the ear phantom, with the measured $L_{\mathrm{eq}}$ value given as a function of time following (a and c) oxygen and (b and d) nitrogen flushing of the cavity. The purpose of these measurements is to show that the recorded gas signals are indeed due to gas on the other side of the membrane (inside the cavity), and not due to spurious gas, and that the water vapor signal, being stable, can serve for normalization purposes. Each data point corresponds to 60 s of signal integration.

As can be observed in Fig. 6(a), the measured oxygen $L_{\mathrm{eq}}$ reaches a high value after flushing with oxygen and stays stable. Then, it decreases gradually to the ambient air equilibrium of 21% oxygen after opening the orifice of the cavity. We note that the oxygen flushing resulted in an initial concentration of about 60%. Similarly, Fig. 6(b) shows that a low signal due to residual oxygen (about 9%) after the flushing with nitrogen is observed, followed by an increase to ambient 21% on the reinvasion of ambient air through the open orifice. Thus, a wide range of oxygen concentrations could be recorded. As indicated with the fitted red curves in Fig. 6, an exponential function with a time constant of about 3 min can be fitted to the experimental data to describe the oxygen gas exchange. We note that middle ear ventilation through the Eustachian tube would in reality be much slower, and in most cases, impractical to monitor. In this study, the ventilation rate of the ear phantom is dependent on the size of the small puncture, and the goal for the measurements of gas exchange is only to demonstrate that the measured signals come from the gas in the tympanic cavity and not due to any spurious gas absorption in the measurement system. A different behavior is observed when comparing Figs. 6(a) and 6(b) with Figs. 6(c) and 6(d), which are the results for the water vapor signal with oxygen and nitrogen flushing, respectively. The $L_{\mathrm{eq}}$ values have no noticeable change after flushing, which can be explained by the establishment of 100% saturated water vapor conditions in the liquid water-containing cavity almost immediately after flushing of the cavity with dry gas. This fact also proves that the water vapor can be used as a reference gas of known concentration to estimate by normalizing the true concentration of the biologically active gas oxygen in the tympanic cavity, as described in Ref. 24. This can then be achieved even if the optical path length is unknown due to scattering, and calibration by ambient concentration oxygen is not available. In addition, the time evolution of the water vapor signal also shows the stability of measurement system. As a matter of fact, it can also be seen from Fig. 6 that the experimental data exhibit some fluctuations during the continuous measurements. This could be the result of still uncompensated residual optical interference fringes, affecting the fitting procedure.

3.2.

Reflectance Spectral Measurements

Ear infection causes the eardrum to change in color from grayish to reddish, and in serious cases, it may lead to a ruptured eardrum. In this study, some human blood was used to cover the surface of the eardrum phantom to simulate an infected eardrum. The measured reflectance spectra from the “normal” (black curve) and the “infected” (red curve) eardrum phantom are shown in Fig. 7. It can be seen that two prominent absorption peaks appear at about 540 and 580 nm, which are the blood absorption imprints. In addition, another peak appears at about 630 nm, which could be the absorption imprint of methemoglobin (Mhb) in the fish swimming bladder.^35,36 For the reflectance spectra of a real human eardrum, the last absorption peak does not exist, so this phenomenon can be ignored in the present study. When the ear suffers from infection, the eardrum becomes red and possibly dilates. The reflectance spectrum from the “infected” eardrum clearly has some differences from that of the normal one, just as the curves indicated in Fig. 7, where the typical two blood absorption peaks are much larger for the “infected” eardrum.

Fig. 7

Reflectance spectra of the membrane in the ear phantom measured by diffuse reflectance spectroscopy. When human blood is added to the fish-bladder wall to mimic a reddened eardrum, the typical hemoglobin absorption imprints at 540 and 580 nm are clearly seen. The differences in spectral shape can be quantified by using an “erythema index,” which in this case differs by a factor of 3.

In order to quantify the erythema of the TM from acquired diffuse reflectance spectra, an erythema index can be defined to quantify the reddening. Such an index was proposed by Feather et al.³⁷ It is evaluated using the blood absorption imprint in the diffuse reflectance and is linearly dependent on the hemoglobin concentration. Index values from the two recorded spectra are indicated in the figure as calculated with a formula, slightly modified from Feather et al.³⁷ We note that the ratio of the index values is about 3, illustrating how the difference in spectral shape can lead to a quantified considerable contrast between normal and diseased conditions. Clearly, the diffuse reflectance spectrum gives more objective and quantitative information than what is possible with human eye inspection, as noted by Sundberg et al.¹⁹