1. Introduction

Mid-infrared (MIR) spectroscopy and spectral imaging are applied in two basic modes: transmission and reflectance. For biological samples, the transmission mode often requires a sample be dissected into thin slices because of the strong absorption. This mode is suitable for in-vitro rather than in-vivo operation, where a sample must be measured in situ. For potential clinical diagnostic applications, in-vivo operation in the reflectance mode is clearly essential.

Interestingly, while MIR spectroscopic imaging has been shown as a valuable analytical tool that can provide simultaneous information on both chemical composition and morphological features of complex materials and structures, extensive collection of important work and reviews of this field in literature [1–7] has been concerned mostly with the transmission/in-vitro mode. MIR diffuse reflectance spectroscopy is used for material chemical analysis, but often without the imaging capability. One likely reason for a lack of development in the reflectance/in-vivo imaging mode is the limitation of enabling component technologies. Typical MIR spectroscopic imaging apparatuses employ broadband incoherent thermal sources for illumination. Thermal light sources have low optical spectral density and brightness. The reflected signal power from biological samples illuminated by a thermal source is often too weak for imaging. Light from a large illuminated area can be collected for more signal power, but this means no spatial resolution for imaging.

A key component technology that can solve this problem is lasers, which can offer several orders-of-magnitude higher optical power, spectral density, and brightness. But MIR lasers hitherto often either lack wide spectral coverage, or are too bulky, complex, and costly. As an example, the demand for MIR synchrotron sources for biological research [8] underscores the need for advanced light sources. But synchrotrons are suitable for research rather than commercial applications. Tunable or multi-spectral MIR semiconductor lasers offer the potential for compact, ease-of-use, and economically cost-effective systems. In a recent work, [9] we performed a system-engineering evaluation of the use of semiconductor lasers for MIR micro-spectral imaging in the transmission mode. The results shown significant advantages of lasers compared with thermal sources.

This paper reports an investigation of laser-based MIR spectral imaging in the reflectance mode for in-vivo applications. A fundamental and critical question is what type of spectroscopic and imaging information can be obtained in this mode, and what system engineering should be done to obtain the most relevant information. There are three key aspects of this question, which are:

i) tissue MIR spectroscopy,

ii) MIR tissue optics, and

iii) system engineering

Tissue spectroscopy is concerned with the fundamental spectral properties, which usually mean the MIR absorption features of the biochemical constituents of a tissue. Mid-infrared tissue optics is concern with the optical propagation in the tissues. And system engineering is concerned with the design and construction of the measurement apparatus to acquire the most relevant spectral and imaging information from a sample.

The first aspect, tissue spectroscopy is intrinsic to the tissue fundamental biological properties and is independent of any specific mode of measurements. The reason for the interest in the MIR spectral range is the richness of the “finger print” fundamental band absorption of basic molecular bonds, e. g. ν_O-H, ν_N-H, ν,σ_(CH3), ν,σ_(CH2), ν_C=O, ν_C-C, ν_PO2, etc., of water, lipids, protiens, nucleic acids, and carbohydrates in tissues. It is in the domain of biological science that is outside the scope of this work, and information can be found in comprehensive reviews. [2,10]

This work aims only for the engineering development of the MIR reflectance imaging technique and is mainly concerned with the other two aspects, which are the MIR tissue optics and system engineering. A comprehensive investigation of these two issues is certainly beyond the length and scope of this paper. This work is an experimental investigation to explore how laser-based measurements may perform in the reflectance mode given typical MIR tissue optics with strong absorption and high inhomogeneity. A more detailed discussion of the issues and approaches in this work is given in the next section.

2. Issues in MIR reflectance imaging through biological tissue and fluid

2.1 Fundamental tissue optics difference between near-IR and mid-IR

There has been extensive work on diffuse reflectance spectroscopic imaging in the near-IR (0.8–2 µm). For examples, the technique has been used to classify skin lesions, [11] to image brain activities, [12] to diagnose the burns, [13] and to study vulnerable plaque of human artery. [14, 15] The theory of diffuse reflection applicable for this wavelength range has been investigated by many authors. [16, 17]

One obvious question is whether all developed features of the near-IR technique can be applied to MIR with only a change of wavelength. A detailed examination indicates that it is critical to recognize that MIR reflectance imaging is of a different nature from that of the near-IR. The difference is due to the fundamental tissue optics difference between the two spectral regions, involving two basic optical coefficients: absorption coefficient µ_a and scattering coefficient µ_s. The absorption coefficient µ_a determines how far light can travel before loosing its intensity while still in its original path, and, the scattering coefficient µ_s determines how far light can travel before loosing its original phase and changes direction (scattered).

Most biological tissues have µ_a in the near-IR much smaller than that of the mid-IR. The reason is the strong MIR fundamental absorption mentioned above. The reverse is true for µ_s. For near-IR, the wavelength is smaller than the typical dimensions of tissue microscopic structures that cause scattering. Thus, Mie scattering is dominant, resulting in large µ_s. On the other hand, the scattering of these structures is Rayleigh-like in the MIR, which scales vs. wavelength λ as λ^-4, resulting in smaller µ_s.

These differences lead to two important distinctions between the techniques. First, in absolute term, owing to a much larger µ_a, the MIR technique is principally for measuring localized and microscopic correlation of tissue morphological features with its biochemical composition. At the other extreme opposite, red and short-wave near-IR penetrates far deeper into tissues. Thus, techniques such as optical diffuse tomography can reconstruct a macroscopic hemoglobin absorption image of a whole organ, for example, but do not reveal any microscopic details of tissue histochemical spectral signatures. Optical coherent tomography was developed for primarily measuring microscopic structural differences between tissues, but less concerned with tissue spectroscopy.

Second, in relative term, while µ_a<<µ_s for near-IR, the reverse is true for MIR. This means that while near-IR light suffers significant phase scrambling long before any significant absorption signature is obtained, MIR light already carries sufficient absorption signature as it propagates long before it suffers significant phase scrambling. This entails that coherent imaging can be applied for MIR over a depth <1/µ_s that is >>1/µ_a. In contrast, coherent imaging is not relevant for near-IR and photon density migration model must be applied. The implication of phase preservation is quite relevant for more advanced techniques such as holography to construct 3D MIR molecular spectral images.

In between, some features in the long-wave near-IR (1.6–2.5 µm) can be similar to either technique, depending on the type of tissues and particular measurement conditions. The key point is that many features that have been developed for the near-IR diffuse reflectance imaging technique are not relevant to the MIR, which requires its own development, and which is the purpose of this work.

2.2 System engineering issues and the approaches of this work

There are two key system engineering issues that are of concerned in this work. One is the issue on the strength of signal, which was discussed above as the reason why thermal sources are inadequate and lasers are more desirable. The second issue is the technique of image acquisition. Figure 1 is used for the discussion of these issues.

Figure 1(a) illustrates a tissue immersed beneath a thin film layer of fluid, e.g. blood, which is actually a good representative of spatially homogeneous tissues with strong absorption. A key challenge of MIR reflectance imaging mentioned above is to be able to see through such an optically thick layer. Figures 1(b) and (c) illustrate the need to distinguish specular surface reflection from the backscattering radiation from a tissue interior, which contains the spectral information of interest. By virtue of the Kramer-Kronig relation, a medium with strong, sharp spectral features also has related spectral features in the Fresnel reflection. However, such spectral features are generally much less pronounced compared with absorption. Figure 1(b) shows an ideal sample that is optically flat (relative to λ). For a spatially uniform illumination, the specular reflection is also uniform and appears as an additive constant on an image, which would easily be filtered. But in general, Fig. 1(c) shows a sample with substantial surface morphological features. Here, it is necessary to discriminate the specular image of the surface morphology from that of the backscattered image of the tissue interior.

 

Fig. 1. Conceptual sketch of the reflectance imaging. (a) illustrates a tissue immersed beneath a thin film layer of fluid (i.e., blood); (b) shows an ideal optically flat sample; and (c) shows a sample with substantial surface morphological features.

Download Full Size | PDF

This work designs a number of exploratory experiments to examine some aspects of these issues. Reflectance imaging test through thin films of water, animal blood, and of biomaterials including plant tissue (leaf) and chicken subcutaneous tissue were performed. Transmittance through water and blood samples was studied for path length of more than 500 µm to test the capability for large dynamic range (several orders-of-magnitude of absorption) with laser power. Owing to the low µ_s, reflectance imaging can be obtained without significant diffusion (blurring) induced by scattering, and a scanning imaging was performed for a target through a 150-µm thick film of blood to study this effect. Finally, MIR reflectance imaging of plant and animal tissues was performed with a coherent imaging system using a staring focal plane array with two laser wavelengths. The images were correlated with the optical images. The rest of paper is organized as follow: Section 3 describes the experimental system and approaches; Section 4 reports the experimental results and interpretations; and Section 5 discusses the potential of laser-based MIR reflectance imaging technique and provides a conclusion.

3. Experimental approaches and setup

3.1 Measurement set-up for signal dynamic range and attenuation coefficient of water and blood

Strong absorption is usually cited as an impediment to MIR in-vivo applications. In order to demonstrate the laser capability to overcome this problem, we first measured the attenuation coefficient of the water and blood over several orders-of-magnitude of absorption, using a variable path-length transmittance assembly formed by two CaF₂ windows. One window is fixed, and the other is mounted on a translation stage with a digital micrometer. The laser beam is sent through the assembly and the transmitted power is measured as a function of the liquid film thickness.

3.2 Scanning imaging system setup

The purpose of the scanning imaging set up is to study the scattering effect through blood. The system focuses the laser beam onto the sample by a two-lens assembly as shown in Fig. 2(a), where a 1″-diameter aspheric ZnSe lens is used to collect and expand the laser beam onto a 3″-diameter Silicon meniscus lens for beam refocusing. The focused laser spot size was determined to be ~50 µm. The backscattered light was detected with an InSb detector. The signal was digitized and processed with a DSP board.

The sample was formed between two CaF₂ windows with a variable optical path length as described above. A target that reflects the laser beam was placed at the far-side window (relative to the incident beam). In between the two windows was either air, water, or animal blood. The sample was mounted on a 2-D translational stage. A computer was used to control the assembly and automatically produce images.

 

Fig. 2. Infrared micro-reflectance imaging setup. (a), scanning imaging system; and (b) staring imaging system. Letter abbreviations in the figure are as follows: L for lens, M for mirror, FPA for focal plan array, S for sample, and W for CaF₂ window.

Download Full Size | PDF

3.3 Coherent imaging with focal plane array

Low µ_s allows coherent imaging as discussed above; and this setup was used to test on biological tissues. The microscopic optical system used here is similar to the one described in a previous study, [9] and illustrated in Fig. 2(b). The microscope employs a single objective lens L2. A CaF₂ 50/50 beamsplitter is used to reflect the laser beam onto the sample. A focal plane array was placed at the image plane of the microscope to capture the image.

3.4 Lasers and detectors

A number of MIR lasers were used for various experiments and listed in Table I. Detectors used in the experiments include a PtSi focal plane array (FPA) (Inframetrics, Bellerica, MA), an InSb photodiode (EG&G Judson, Montgomeryville, PA) and a MCT photodiode (Kolmar Technologies, Newburyport, MA).

4. Experimental results

4.1 Measurement of dynamic range and attenuation coefficient of water and blood

Deionized water and blood collected from a Watanabe rabbit were used for the measurement. Anti-coagulation agent was added immediately after the blood was collected, and the sample was stored at 4°C.

Table I. Laser characteristics

View Table

The results for water and blood at 5.4-µm wavelength are shown in Fig. 3(a) and 3(b), respectively. Attenuation coefficients were obtained from the measurement. Figure 3(c) compared the results obtained here with the published water attenuation coefficient from Ref. 18. Excellent agreement between the two can be seen. Most importantly, over 6 orders-of-magnitude of absorption can easily be measured, thanks to the high laser power. This means that laser photons can be transmitted through highly absorptive film of water and blood, back-reflected from a tissue, and detected with sufficient signal-to-noise for imaging purpose. Yet, the laser power level used in this experiment, as listed in Table I, is still very modest compared with state-of-the-art, and the power density was still several orders-of-magnitude below typical hyperthermia necrosis limit. [19]

 

Fig. 3. Experimental results for attenuation coefficients of water and blood. (a) and (b) Measurements (symbols) and least-square fit (line) of relative transmitted power as a function of film thickness for water and animal blood, respectively (wavelength 5.4 µm); (c), Comparison of the measured attenuation coefficients at different wavelength to those in Ref. 18, demonstrating excellent agreement. (d), the difference between blood and water attenuation coefficient at various wavelengths, showing the effect of blood absorption in the fingerprint region.

Download Full Size | PDF

The difference of attenuation coefficients between water and blood is plotted in Fig. 3(d). It can be seen that attenuation coefficient of water and blood is almost the same at 4.6 and 5.4 µm, while differs significantly at other three wavelengths. This can be explained based on the constituents of blood. Blood consists of complex liquid plasma in which the cellular elements, erythrocytes (red blood cells), leukocytes (white blood cells), and platelets (thrombocytes), are suspended. The plasma contains 90% of water and forms more then 55% of blood volume. The rest of the blood volume is almost all occupied by red blood cells. Hemoglobin, the protein molecule forming the red blood cells, has little absorption at 4.6- and 5.4-µm. Therefore, the attenuation is mainly the water contents of the blood. In fact, a discernible lower attenuation coefficient of blood compared with water at these wavelengths may indicate the lower water volume in blood. In the 7–10 µm regions, hemoglobin has a stronger absorption than water, contributing to a larger total attenuation coefficient as evidenced in the figure.

4.2 MIR imaging of through a thin film of blood

As discussed above, an implication of the high laser power advantage is the ability to image through a thin film of blood. In addition, low scattering coefficient µ_s allows imaging without significant diffusion. This experiment was designed to study these issues.

 

Fig. 4. Infrared micro-reflectance imaging of a gold-coated patterned sample through a thin layer of water or blood at 5.4 µm. (a), a white light image of the sample, notice the alignment marker in the bottom of the image; (b), the IR laser image without liquid above the sample; (c), a visible-spectrum picture of the sample with a thin blood film on top, only the blood can be seen; (d)-(f), the IR laser images with a thin layer of water above the sample, the water thickness is 50-, 100-, and 150-µm from left to right. (g)-(i), the IR laser images with a thin layer of blood above the sample, the blood thickness are the same as those of water. Little difference between water and blood images indicates the lack of significant scattering by particles in blood at this long wavelength.

Download Full Size | PDF

The image target is a patterned line marks on a gold-coated surface as shown in Fig. 4(a), which was obtained under white light. The spacing between the lines is 500 µm, and the total image area is about 2.5×2.5 mm². The target was immersed under a thin film of water or blood and inserted in the 2-D scanning imaging set up described in Section 3.2. Figure 4(c) is a picture showing the target under a thin film of blood, which was not visible due to the strong scattering of blood in visible light. Figures 4(d-f) are the images of the sample under a thin-film of water, with film thickness of 50, 100, and 150 µm respectively. Figures 4(g-i) are the corresponding images with a thin film of blood having the same thicknesses as those of water. The MIR images was obtained for 5.4-µm wavelength, where both water and blood have similar attenuation coefficients to ensure comparable signal-to-noise levels and do not affect the image comparison.

Comparison of water (no scattering) with blood images obtained clearly demonstrated that the scattering of blood is so low that it can be neglected. The fact that the system can “see” through 100-µm of blood at 5.4 µm without significant image “smearing” suggests that coherent imaging is possible with MIR lasers.

4.3 Multi-wavelength MIR coherent imaging of biological tissues

Blood is the simplest biotissue consisting of a single dominant protein, hemoglobin, and the plasma media of which water is the primary component. The other tissue types, generally speaking, are much more complicated both in morphology and chemistry. It is beyond the scope of this paper to explore the details of the different tissue types and their images. The purpose of these experiments is to illustrate the potential of MIR laser imaging with a simple coherent imaging technique for a few selected samples, including a house plant leaf and an inner chicken skin.

4.3.1 Image of a plant leaf

The results for a leaf from a house plant (Epipremnum Pinnatum) are shown in Fig. 5(a). For the plant leaf, the inner tissues are surrounded by the upper and lower epidermis. Immediately beneath the upper epidermis is a parallel array of columnar cells of palisade mesophyll. The loosely arranged sponge mesophyll cells are networked with veins that consist of elongated cells. [20]

 

Fig. 5. Infrared micro-reflectance imaging of a house plant leaf (Epipremnum Pinnatum). (a), a white-light picture of a leaf; (b), a white-light image of a small section of the leaf, notice the upper epidermis and the vein; (c), MIR laser image of a small section of the leaf taken with the staring imaging system; (d), MIR laser image of the same section when the depth of focus is changed, showing an image from a deeper layer from the upper epidermis, which would not be visible with white light.

Download Full Size | PDF

As shown in Fig. 5(b), the visible-spectrum image of a portion of the leaf demonstrated the upper epidermis with the cell structure clearly shown. In the picture, one can also find a clear indication of a vein underneath, with the epidermis cells forming elongated patterns under the influence of the vein cells. Figure 5(c) is the image illuminated with a 4.6-µm laser operating at 10% duty cycle and an average power of 2.5 mW. When the optical system focused on the leaf surface, an image similar to that of white light was obtained, showing similar epidermis cell morphology. The polygon shaped epidermis cell as well as the elongated above-vein type cell can all be seen under laser illumination. However, as the optics was focused onto a plane beneath the leaf surface, a very different image was obtained as shown in Fig. 5(d). This image is impossible to be seen with visible light, because of the strong upper layer scattering. We tentatively assign this image as from the deeper layer of the leaf since this is consistent with the fact that the epidermis cell walls are curved where they meet and likely to have higher reflectivity than the membrane surface. There was no effort to dissect the leaf and examine the origin of this image, as this experiment is only interested in exploring the applicability of coherent imaging at this wavelength to study tissues with layers. This experiment suggests that confocal techniques can be used to image a tissue at different depths.

4.3.2 Image of animal tissue through a lipid layer

An equivalent experiment was conducted on a small section of chicken skin as shown in Fig. 6(a). The internal structure of chicken skin is generally similar to that of other animal skins. Fresh chicken skin can be divided into two main layers-the surface growing layer or epidermis, and the deeper layer of relatively unchanging connective tissue, called the corium or dermis. Below the dermis in the live bird is the subcutis or subcutaneous layer composed of loose connecting tissue that contains blood vessels, nerves, and fat cells. [21] The visible image (Fig. 6(a)) shows the subcutaneous layer with blood vessels beneath a thin fatty layer.

 

Fig. 6. Infrared micro-reflectance imaging of the micro-blood vessels from chicken skin. (a), the white light image of a small section of the chicken skin, notice the micro-blood vessel shown in the image; (b), the image of the surface using 3.4 µm laser illumination, the blood vessel was not seen; (c), the MIR laser image when the focus depth is changed to a subsurface plane, showing the blood vessels; (d) the same as (c) but using 4.6 µm laser; (e), the false color composite image both images (c) and (d).

Download Full Size | PDF

Two wavelengths were used in order to demonstrate multi-wavelength spectral imaging. The two wavelengths, 3.4 and 4.6 µm are not in the fingerprint region, but they were used because they are within the spectral response of the PtSi FPA. Images of different foci at different depth in the tissue were obtained similarly to the experiments with the plant leaf above. The images at the surface layer (a thin layer of lipid) reveal no blood vessels structure as shown in Fig. 6(b) under the 3.4-µm laser illumination. Only lipid and some connective tissues can be seen. However, at a proper focal depth, images of the blood vessels came into the focus. By comparing Figs. 6(b) and (d), which are two 3.4-µm images taken at different focal depths, it has demonstrated that subsurface morphological and spectral information can be obtained. It is also shown that the blood vessels in the Fig. 6(d) are the same as that seen in the visible-image in Fig. 6(a). The same focal depth effect was observed for the 4.6-µm image. The focused images of both wavelengths are shown in Figs. 6(c and d).A false color image is constructed by fusing the two different-wavelength images, with 3.4-µm image represented as red, and the 4.6-µm image as green. The resulted Fig. 6(e) shows excellent intensity correlation between the two images, and offers a more vivid composite to be compared with the visible-spectrum image.

Clearly, the two wavelengths used here are far from optimal to extract meaningful spectroscopic information. However, the purpose here is only to demonstrate the potential and feasibility of the MIR in-vivo spectral imaging technique, not the spectroscopy of chicken skin. It is conceivable that a system with many tunable lasers covering a wide spectral range, combining with a corresponding FPA in the fingerprint spectral region can produce more biologically meaningful spectral images. Algorithm can be developed for image fusion such that it produces spectroscopically relevant color-coded images to enhance and/or contrast a tissue histochemistry that are useful to clinicians for diagnostics.

5. Discussion and conclusion

This work is an effort to evaluate the feasibility of semiconductor-laser-based MIR reflectance imaging, which can be used for in-vivo applications. In the transmission mode, MIR micro-spectroscopic imaging has been shown highly valuable for many applications, but it is a challenge to apply the technique to the reflectance/in-vivo mode because of the weakness of optical signals owing to the strong absorption. This work shows that lasers of modest power can overcome the large absorption loss and perform reflectance imaging to a depth of several 10’s if not 100’s of µm of thin layer of biotissues (i.e., blood, lipid). The high power, brightness, and low noise feature of lasers may allow real-time or nearly real-time imaging that is potentially useful for in-vivo diagnostics.

This work also examines some fundamental and unique aspects of MIR spectral imaging compared with shorter wavelength with regard to tissue optics. As expected, long wavelength light suffer very little scattering compared with absorption, which is opposite to that of short wavelength. Thus, MIR spectral imaging can be performed with coherent imaging. Although outside the scope of this work, many conventional optical engineering techniques can be used to help discriminate surface morphology image from the tissue internal absorption image to enhance the image fidelity, thanks to the lack of scattering. For example, confocal technique clearly can be used to enhance both planar and depth resolution for 3D imaging. Polarization discrimination techniques [22, 23] can also be applied. Clearly, more extensive theoretical and experimental work on backscattering, e. g. a study of Mueller scattering matrix for a wide range of tissue types, can achieve further understanding and develop more optimal techniques with both high spectral and image fidelity that are suitable to any specific tissue of interest.

A large number of wavelengths are needed to fully exploit the power of MIR spectral imaging. As mentioned earlier, the lack of spectral coverage has always been an issue with lasers, and it may not be cost-effective to have a large number of fixed wavelength lasers to cover the entire region. However, it is conceivable that a modest number, e. g. ~5–10 broadly tunable lasers [24, 25] can be combined via the wavelength-division multiplexing technique to provide most of spectral coverage requirements. Since semiconductor lasers are compact and simple to operate, the integration of these lasers will not likely have the problem of bulky hardware and complex control that has also been a concern for many large laser systems.

In conclusion, in regard to both system engineering and tissue optics, laser-based MIR reflectance spectral imaging technology is quite feasible for in-vivo biological applications. Developing such a tool will enable biomedical research that may eventually lead to useful clinical diagnostic applications.