1. Introduction

Optical spectroscopy has been shown to discriminate early cancers by detecting alterations in the optical properties of human epithelial tissues, where 80% of cancers originate [1–3]. These changes result from biochemical and morphological modifications that accompany disease progression such as increased nuclear size, hyperchromasia, pleomorphism, and increased metabolic activity of epithelial cells [4]. Precancer development also involves the underlying stroma, with loss of collagen fiber density owing to release of matrix metalloproteases (MMPs), as well as an increase in microvascularization stimulated by angiogenic factors such as vascular endothelial growth factor (VEGF) [5–11]. The depth dependent morphological, architectural and biochemical alterations that occur with precancer make interpretation and modeling of spectroscopic signals challenging, as the typical measured optical spectrum is a mixture of signals from a range of depths. For example, the increased scattering in epithelial tissue during precancer progression can be difficult to detect because of the concomitant decrease in scattering from the stroma [7, 12–17]. Analogously, two opposing processes can obscure detection of the origins of fluorescence: increased fluorescence from the reduced form of nicotinamide adenine dinucleotide (NADH) in the epithelium and decreased stromal fluorescence [7, 18–20]. Further, it has been demonstrated that the optical scattering and fluorescence properties vary with depth within the epithelium with progression of dysplasia [7, 16, 21]. These depth variations can be correlated with different stages of carcinogenesis or benign conditions such as hyperplasia or inflammation. Therefore, depth resolved optical spectroscopy of epithelial tissue would greatly aid the isolation of optical signatures associated with carcinogenesis and as a result could improve the noninvasive detection and monitoring of precancers.

Endoscopic probes with fiber-optic elements provide the possibility for examination of tissue optical properties at different locations within the body [2, 22]. Widely used for its simplicity are flat-tipped fibers (FTF). It has been shown that this design can provide depth selectivity to some extent with probes that utilize multiple source-detector separations [23–25], or a single-fiber geometry using a variable aperture technique [25, 26]. However, these design solutions lack the ability to effectively isolate the relatively small epithelial optical signal from the strong stromal component. Differential path length spectroscopy showed promising results localizing the optical properties of the most superficial layer of tissue [25–27]. This approach was used to extract the optical properties of tissue with very thin epithelial tissue, such as in the breast and bronchial tree [28, 29], however measurements in the oral cavity were more challenging because of variable epithelial thickness and keratinization [30]. Recently, encouraging results for depth resolved measurements targeting epithelial tissue were obtained using probes with a ball lens [31–33] or with obliquely oriented fibers [34–36]. It was shown that ball lens designs can efficiently localize the optical signals close to the probe tip, however, miniaturization of the probe is limited by the size of the ball lens. Additionally, the manufacture of a ball lens assembly is relatively complex, requiring precise alignment of fibers with respect to the center of the ball lens as well as careful consideration of the specular reflection losses due to the high refractive index of sapphire - the material commonly used for ball lenses. Previously, we proposed and evaluated a fiber optic probe with obliquely oriented collection fibers to increase both depth selectivity and collection efficiency of the scattering signal from epithelium in oral cavity tissues [34, 37, 38]. Initially, we demonstrated that the probe can be used to detect scattering from a specific depth using three-dimensional phantoms that consisted of either polystyrene beads or epithelial cells and collagen [34]. The fiber optic probe was then evaluated in a pilot clinical trial to detect precancer in the oral cavity [38]. The oblique fiber design was found to be sensitive to four diagnostic categories: normal, benign, mild dysplasia, and high-grade dysplasia plus carcinoma. Encouragingly, separation of visually indistinguishable diagnostic categories was achieved with good sensitivity and specificity. For example, discrimination of benign lesions from high grade dysplasia and carcinoma was accomplished with a sensitivity of 100% and a specificity of 85%, while separation of benign from mild dysplasia gave a sensitivity of 92% and a specificity of 69% [38]. These results demonstrate that reflectance spectroscopy using obliquely oriented fibers has the potential to aid screening and diagnosis of epithelial precancers. However, a drawback of this design is that the probe was intended to collect optical signals only from a fixed predetermined depth in the tissue and, therefore, did not provide information about changes in tissue optical properties with depth. This complicates analysis of in vivo data due to variation in epithelial thickness that can be associated with conditions like hyperplasia and keratinization. Additionally, miniaturization of the probe design is limited by the long-term bending radius of the fibers (approximately 300–400 times the fiber diameter), which is the minimum radius of curvature that an optical fiber can bend long-term without significant signal loss or mechanical failure.

Here we describe a compact probe design optimized for the collection of spectroscopic signals from multiple depths in tissue. This design incorporates multiple beveled collection fibers (BF), a flat-tip illumination fiber, and an additional single flat-tipped collection fiber. All fiber axes are oriented normal to the tissue surface, thus eliminating the need to accommodate the long-term bending radius that was used in the previous design [34, 38]. The additional FTF collection fiber was used to collect signals originating more deeply in tissue and also to provide a comparison benchmark in order to evaluate the performance of the BFs. The dependence of collection efficiency and depth resolution on bevel angle and source-detector separation was assessed using a thin white substrate, a multilayer tissue phantom simulating the scattering properties of precancerous tissue, and, finally, on normal human oral mucosa in vivo. Our results demonstrate that beveled multi-fiber probes are an effective tool to perform depth-resolved spectroscopy, which has the potential to improve precancer detection and monitoring.

2. Materials and methods

2.1 Probe design and fabrication

Figures 1(a) and 1(b) illustrates two collection fiber arrangements that were used in this study. An air gap between the beveled fibers and the protective fused silica window takes advantage of the large refractive index change to deviate the collection acceptance cones, providing an overlap between the illumination and collection cones at different depths within tissue. In principle, fibers closest to the illumination beam sample the superficial-intermediate layer of tissue, and those fibers farthest from the illumination beam sample deeper tissue regions. An additional benefit of orienting the fibers along the same axis, as opposed to an oblique orientation, is that it avoids the fabrication complication posed by polishing obliquely oriented fibers where the source-detector separation is dependent on the amount of polishing.

In Figs. 1(a) and 1(b), the illumination FTF (shaded yellow) delivers broadband light from the source (Ocean Optics, HL2000HP-FHSA) in the wavelength range 400 – 900 nm to the tissue or sample region of interest. The adjacent FTF and the BFs collect the remitted photons. All fibers are in direct contact with their nearest neighbor and have a core/clad diameter of 100/110 microns and 0.12 NA (CeramOptec). The protective polyimide jacket was removed to minimize the center-to-center fiber separation. The fiber NA was chosen so that the axial extent of the overlap of the illumination beam and collection acceptance cones would be as small as possible for probing of sub-layers within the epithelium.

The first step in the probe fabrication was alignment of the optical fibers along the same plane. Bare optical fibers were laid flat on a microscope slide and held in place with double-sided tape. A small amount of low viscosity biocompatible epoxy (Epo-Tek 301-2) was applied to the fibers. Capillary forces wicked the epoxy between the fibers. After the epoxy hardened, the assembly was bathed in ethanol to remove the fibers from the microscope slide. Then, the fiber ribbon was inserted into a stainless steel cylinder 27 mm long that had a through slot cut into it. The stainless steel cylinder was beveled to the desired angle prior to inserting the fibers. The end of the fiber ribbon extended beyond the end of the steel cylinder approximately 1 cm. Epoxy was added to secure the fiber ribbon inside of the cylinder. After the epoxy hardened, the fiber excess was cleaved and polished to a 0.1 μm finish using a custom made fiber polishing puck. The polishing procedure was performed on the beveled face and then to the FTF surface. The final polished distal end was inserted into an annealed stainless steel (316L) tube 4.8 mm in diameter and 30 cm long. Using the biocompatible epoxy used in the previous steps, the fiber assembly and steel cylinder were secured inside of the steel tube. Then, a glass window ca. 160 μm thick was placed on the distal end to protect the fibers and to maintain an air gap. Figure 1(c) is a picture of the constructed probe in a fluorescing medium. Light is channeled through the FTF illumination fiber and three BFs to illustrate the overlap of the illumination and collection cones. Figure 1(d) is a picture of the final probe.

 

Fig. 1. Fiber probe and spectroscopic system. (a) & (b) Illustration of probe design utilizing beveled fibers for sampling multiple tissue depths. C1 – the arrangement used to perform white substrate experiments. C2 – the arrangement used to perform phantom and in vivo measurements. Yellow color highlights the illumination fiber and illumination cone. (c) Picture of constructed probe in fluorescing medium with light channeled through the illumination fiber and three beveled fibers. (d) Picture of the final constructed probe. (e) Overall system schematic: 1 – probe, 2 – FC/APC connectors, 3 – light source, 4 – coupling probe, 5 – imaging spectrometer.

Download Full Size | PDF

2.2 System description

Figure 1(e) shows the overall system schematic. This set-up is modular permitting testing of multiple probe designs. The proximal ends of the collection fibers are connected to a coupling fiber bundle (RoMack, Inc.) via FC/APC adapters. FC/APC terminators were chosen for their low insertion loss (ca. 0.15 dB). The fibers of the coupling fiber bundle (100/110 core/clad, 0.22 NA) were stacked in a vertical array using a slit ferrule at the proximal end and aligned with the entrance slit of the imaging spectrograph (PI Acton SpectraPro SP-2356, Pixis 2KB). The triple grating spectrograph was used with either a 150 g/mm or a 300 g/mm grating optimized for visible wavelengths (500 nm blaze) with a theoretical spectral resolution of 0.714 nm and 0.353 nm, respectively. The spectrograph dispersed the light from each fiber onto the imaging CCD. The image produced by the CCD has the vertical spatial dimension along the y-axis and the wavelength dimension along the x-axis. With this arrangement, the spectra from all collection fibers can be acquired simultaneously in a single image. After data acquisition, the image was masked into multiple strips, where each strip corresponded to the position of a fiber at the entrance slit. Each strip was then binned vertically to yield the intensity versus wavelength spectrum. Dark subtraction and post-processing was performed with MATLAB® software.

2.3 Probe characterization

The depth selectivity of the fiber optic probe was characterized using a diffuse white scatterer that was composed of white Teflon tape (90 μm thick) affixed to a thick glass substrate. This sample modeled the ideal case of an infinitely thin diffuse scatterer.

The sample was placed on a translation stage to control the probe-sample separation and the probe was fixed in a holder directly above it. The separation between the probe and the sample was controlled using a fine-pitched micrometer. The initial distance of the probe from the sample was set at 3 mm, which was the largest distance used for probe-sample measurements. Water was added between the substrate and the probe to reduce the refractive index mismatch between the probe tip and the sample surface, thus simulating an in vivo measurement in which saliva is between the probe and tissue. Using the micrometer, the substrate was gradually brought closer to the probe. For each fiber, the signal intensity (integrated over all wavelengths) was measured at each probe-sample separation. Probe assemblies with three bevel angles were evaluated: 35, 40, and 45 degrees (relative to the window surface). For each bevel angle, the depth profiling measurement series were performed in triplicate.

2.4 Phantom preparation

In order to evaluate the depth sectioning ability of the optimal probe design in vivo, signal intensity measurements were taken on multilayer phantoms mimicking the optical scattering properties of precancerous epithelial tissue. A three layer phantom was prepared using polystyrene beads (BangsLabs, Inc.) embedded in 3% w/v agarose (Gibco BRL Ultrapure Agarose). The top, middle, and bottom phantom layers contained beads with size distributions of 5.01 ± 0.14 μm, 8.31 ± 0.66 μm, and 2.50 ± 0.16 μm, respectively. The manufacturer specified refractive index of the beads was 1.59 and the refractive index of the agarose was measured to be 1.335 with a refractometer (Bausch and Lomb). The beads were diluted from stock solution to give scattering coefficients approximating normal epithelial (top layer, ca. 33 cm^-1), precancerous (middle layer, ca. 71 cm^-1), and stromal tissue (bottom layer, ca. 189 cm^-1)[21, 39, 40]. Bead concentrations were calculated using the web-based Mie Scattering Calculator by Scott Prahl (http://omlc.ogi.edu/calc/mie_calc.html). First, the bottom layer was fabricated by pipetting the liquid bead-agarose mixture into a cylindrical well 12 mm in diameter and 6 mm deep. The cylindrical well was machined into an aluminum slab ca. 10 cm × 3 cm × 3 cm. A glass microscope slide was placed atop the phantom and flush against the aluminum slab to ensure a flat upper surface. After the bottom layer solidified, the glass slide was removed and two no.1 glass coverslips approximately 160 μm thick were placed on the top surface of the aluminum slab on either side of the cast bottom layer. The coverslips acted as spacers defining the thickness of the middle phantom layer. A second bead-agarose mixture was prepared to form the middle layer and pipetted onto the upper surface of the bottom phantom layer. A microscope slide was again placed with gentle pressure atop the new layer and flush against the coverslips. After the middle layer solidified, the top layer was then formed repeating the above procedure.

Single layer phantoms, 6 mm thick, were also constructed for measurement of the pure scattering spectra from each layer. After the spectroscopic measurements, phantoms were transversely sliced into approximately 200 – 500 μm thick sections with a Krumdieck tissue slicer (Alabama Research and Development) and images were acquired with an optical microscope (Leica Microsystems, DM6000 M) in brightfield transmittance mode in order to characterize the morphology of the phantoms.

2.5 Phantom measurements

Measurements were taken in a similar manner as described for the thin white substrate, but with the illumination and FTF switched as shown in Fig. 1(b). Our reasoning for switching the FTF fibers was based on the results of the white substrate measurements (shown in section 3.1) where we observed the most superficial sampling to be from fibers closest to the source fiber. In the white substrate measurements, the bevel fiber closest to the source (BF1) was approximately 110 μm from the source (edge-to-edge). We anticipated that a collection bevel fiber with zero separation from the source (BF0) would have improved superficial sampling. Therefore to explore the full range of source detector separations, we switched the illumination and collection FTF to the C2 geometry shown in Fig. 1(b) for the subsequent phantom and in vivo measurements.

Three independent depth profiling measurement series were performed for each phantom type – single layer phantoms and the three-layer phantom. The scattering spectrum collected by each fiber was dark subtracted and normalized by the same fiber’s spectrum from a white reflectance standard (Labsphere, SRS-99). All spectra were converted to wavenumber space for Fourier analysis. A Hann window was used to smooth the spectrum at both ends of the spectral range to prevent aliasing artifacts. Using the FFT function in MATLAB®, a fast Fourier transform (FFT) was taken of the scattering spectra of each single layer phantom to identify the primary frequency components in the isolated layers. Next, these primary frequency peaks were identified in the FFT of the three-layer phantom. Then, the amplitude of the primary FFT frequencies of the three-layer phantom was plotted as a function of probe-sample separation for each fiber.

2.6 In vivo measurements

The 40 degree bevel probe was evaluated in vivo on oral mucosal tissue of a normal volunteer. With the volunteer’s consent, the probe was placed in direct contact with either the inner portion of the lower lip or the dorsal tongue. Four sites were measured for each anatomical location. The resultant reflectance spectra were normalized to one at 610 nm to allow comparison of the relative hemoglobin absorption measured by each collection fiber.

 

Fig. 2. Depth profiling curves. Total intensity of light scattered from a thin white substrate integrated over all wavelengths plotted as a function of distance from the probe tip for probe bevel angles: (a) 35 degrees, (b) 40 degrees, (c) 45 degrees. Abbreviations: FTF - flat-tipped fiber, BF - beveled fiber. Numeration of the beveled fibers corresponds to numbering shown in Fig. 1 (a).

Download Full Size | PDF

3. Results

3.1 Probe characterization

The goals of this probe design are to isolate signals within a few hundred microns from the probe tip, to measure spectra from multiple depths (shallow, intermediate, and deep) simultaneously, and to have high collection efficiency. We used these design criteria to compare three bevel angles: 35, 40, and 45 degrees. We compared the depth profiling curves (total integrated intensity vs. probe-sample separation) plotted in Fig. 2. A peak in a curve indicates the distance from the probe tip, and hence the approximate depth into tissue, from which the scattering contribution will be greatest. BF1 corresponds to the beveled collection fiber with a source-detector separation of 110 μm. Each subsequent BF is separated by an additional 110 μm. For all bevel angles evaluated, Fig. 2 shows that BFs nearest to the source fiber interrogate most shallowly, while BFs farthest from the source fiber interrogate more deeply.

An important consideration when using flat-tipped fibers with small source-detector separations is the signal contribution due to specular reflection from the fused silica window. This reflection artifact is evident as a nonzero plateau in the depth profiling curve for the FTF in Fig. 2. The specular signal can be minimized by i.) polishing the FTF at a slight angle; ii.) using an anti-reflection coating on the glass window, and iii.) subtracting the measured probe signal from a black substrate. For the purposes of this paper, the above procedures were not implemented in order to compare the specular reflection component for all collection fibers. For the beveled fibers it was found that this component is less than 4% of the peak signal.

 

Fig. 3. Comparison of bevel angles. (a) Distance from probe tip that corresponds to the maximum signal intensity in the depth profiling curves shown in Fig. 2. (b) FWHM of depth profiling curves for each collection fiber. (c) Maximum signal intensity for each beveled fiber relative to the flat-tipped fiber (FTF - flat-tipped fiber, BF - beveled fiber).

Download Full Size | PDF

Two metrics were used to compare the depth resolution of probes with different bevel angles: the distance from the probe tip that has maximum signal intensity and the full width at half maximum (FWHM) of the depth profiling curves shown in Fig. 2. The range of depths from which the collection fibers gather scattering signal depends on the amount of overlap between collection acceptance cones and the illumination cone, and differs according to bevel angle and source-detector separation. The maxima of the depth profiling curves and corresponding standard errors for each fiber are shown in Fig. 3(a). All fibers exhibit deeper penetration with the 35 degree bevel as compared to the 40 and 45 degree bevels, which interrogate more shallowly. The data shown in Fig. 3(a) were adjusted such that the average probing depth for the FTF fiber would be the same for all bevel angles.

This adjustment accounts for the experimental error in identifying the point where the probe and sample are in direct contact. The optimal probing depth of the FTF was found to vary no more than 20% from the average of all three bevel angles.

The FWHM of the depth profiling curves and corresponding standard error for each collection fiber is shown in Fig. 3(b). The FWHM increases for all fibers with increasing source-detector separation, which is consistent with the diverging of the illumination and collection acceptance cones. The 35 degree beveled fibers have the largest FWHM for all fibers, and therefore the poorest depth resolution, while the 40 and 45 degree beveled fibers demonstrate better depth resolution.

Comparison to the FTF shows that in all cases the beveled fibers closest to the source fiber have shallower probing depths and significantly better depth resolution. For example, the average optimal probing depth for the FTF is 716 μm, which is two to three times larger than the optimal probing depth for BF1. Likewise, the average FWHM for the FTF is approximately 3 times larger than the FWHM for the 35 degree BF1 and 5 times larger than the 40 degree BF1.

 

Fig. 4. Three-layer phantom. (a) 20x transmittance brightfield image of a transverse cross-section through three-layer bead phantom. (b) Scattering spectra of single layer bead phantoms corresponding to the top, middle, and bottom layer of the three-layer phantom plotted as a function of wavelength.

Download Full Size | PDF

Normalization of the collected signal intensity with respect to the FTF is shown in Fig. 3(c). Of note is the superior collection efficiency of the BFs closest to the source fiber. For BF1, the collection efficiency is a factor of 3.3, 2.9, and 1.3 greater than the FTF for bevel angles 35, 40, and 45 degrees, respectively. For all bevel angles, the collection efficiency drops rapidly with distance from the source fiber. The decrease in the relative intensity for 45 degree BFs is dramatic. The reasons for this are foreshortening of the collection aperture relative to the illumination area and losses from Fresnel reflections at the air-fiber interface.

The bar graphs in Fig. 3 illustrate the trade off between signal intensity and depth localization. Although the 45 degree BFs sample most shallowly and have the narrowest FWHM, the poor collection efficiency and the substantial overlap of the BF profiles excludes it from the final probe design. In contrast, the 35 degree BFs have the best collection efficiency, but at the expense of depth resolution (relatively high FWHM) and the ability to isolate scattering from the most superficial region (compare optimal probing depth of 683 μm for the 35 degree BF1 to 367 μm for the 40 degree BF1). Therefore, we chose the 40 degree bevel for the final design as it combines the best features from 35 and 45 degree bevel: shallow interrogation, high collection efficiency, and good depth isolation.

3.2 Phantom experiments

The probe with the 40 degree bevel was evaluated using a multilayer tissue phantom with scattering properties mimicking normal and precancerous epithelial tissue and stroma. Figure 4(a) shows a transverse cross-section through the three-layer phantom imaged in transmittance bright field. The beads embedded in the top and middle layers are plainly visible in the image, while the bottom layer is almost opaque owing to the high concentration of beads required to reproduce the scattering coefficient of stromal tissue. The top and middle layers are ca. 230 and 260 μm thick, respectively, and the bottom layer is ca. 6 mm to replicate a semi-infinite layer. The three layer configuration simulated the development of mild to moderate dysplasia where basal cells proliferate, encompassing the bottom 1/3 to 2/3 of the epithelium [4].

The beads used in our phantom fell within the size range 2-8 μm, resulting in angular scatter that is strongly forward directed. The highly forward scattering nature of epithelial tissue has been demonstrated in several organ sites. Mourant, et al. [41] has measured the anisotropy factor (g = <cosθ>) of tumorigenic and nontumorigenic mammalian cells to be very highly scattering at 0.98 for 633 nm light. Anisotropy of bladder tissue has been measured in the range 0.85–0.92 at 633 nm [42]. Anisotropy values for epithelium and stroma in the range 0.94–0.97 and 0.8–0.89, respectively, have been used in theoretical modeling of reflectance measurements for light in the visible wavelength range [33, 36, 43]. The phantom described in this manuscript has anisotropy values of: top − 0.89, middle − 0.92, bottom − 0.83, which are consistent with literature values.

For the three bead sizes used in the phantom study, the top, middle, and bottom layer reduced scattering coefficient was 3.6 cm^-1, 5.7 cm^-1, and 32.1 cm^-1, respectively, which conforms with literature values. The top and middle layer μ_S′ values are consistent with measurements at 633 nm by Ramachandran, et al. [44] using tumorigenic and nontumorigenic mammalian cell model of tissue, where the μ_S′ was found to be in the range 3 – 4.5 cm^-1. Measurements of minced homogenized human oral biopsies by Muller, et al. [45] have obtained higher μ_S′ values of 10–30 cm^-1 extracted from diffuse reflectance spectra (360–700 nm). In their study, oral tissue sections from the epithelium and submucosa were homogenized using a mortar and pestle. Since the stroma, or submucosa, occupies a larger volume of tissue than the epithelium, it can be assumed that the high μ_S′ values obtained are dominated by the stromal layer of tissue. Our phantom optical parameters fall within these currently published values providing a good first approximation of scattering in epithelium and underlying stroma

Initially, scattering spectra were collected from the single layer phantoms which had the same composition as the layers shown in Fig. 4(a). It is evident from Fig. 4(b) that the scattering spectrum from each individual layer has a characteristic ripple frequency associated with the bead size that was used in preparation of the layer. This characteristic ripple frequency provides a unique identifier for the analysis of the depth profiling capability of the probe under scattering conditions analogous to those found in biological tissue. In Fig. 4(b), the curves are smoothed using a 3.5 nm window for clarity, but were kept in their original form for data analysis.

The single layer spectra are plotted as a function of wavenumber in Fig. 5(a). A fast Fourier transform (FFT) was taken of the measured spectrum (in wavenumber space) for each layer alone and was used to identify the primary frequency components in the original scattering spectrum. Figure 5(b) is a plot of the FFT power spectrum from the single layer phantoms illustrating the characteristic frequencies associated with the three different bead sizes. Simulated Mie spectra were also generated based on the manufacturer bead specifications. The primary frequencies from the FFT of the simulated Mie spectra were consistent with the values obtained from the measured scattering spectra.

Figure 6 is the measured scattering spectrum (in wavenumber space) from the three-layer phantom at zero probe-sample separation. Qualitatively, the characteristic bead ripple pattern for the top layer is visible in spectrum from BF0. The middle layer ripple pattern stands out at small wavenumbers for the BF1 spectrum, while the bottom layer ripple pattern becomes more prominent for spectra obtained by BF2-BF5. Given the primary frequency components of each layer obtained from Fourier analysis of the single layer phantoms, it was possible to track the contribution of a given layer (top, middle, or bottom) to the total collected scattering signal as a function of distance from the probe tip. Figure 7(a) is a plot of the FFT amplitude of the respective characteristic frequency components for the top, middle, and bottom layer as a function of probe-sample distance. Beveled fibers significantly outperform the flat-tipped fiber in terms of signal collection efficiency and depth resolution when the probe is in direct contact with the phantom. Similar to the white substrate measurements, the BFs closest to the source probe most shallowly and those farthest from the source probe more deeply. Figure 7(b) emphasizes this point, with an illustration of the phantom above the depth profiling plots for the C2 fiber configuration that was used in three-layer phantom measurements (1Fig. 1(b)).

 

Fig. 5. Primary frequency components of single layer phantoms. (a) Scattering spectra of single layer bead phantoms plotted as a function of wavenumber. (b) Fast Fourier Transform of single layer phantom spectra showing the characteristic frequencies associated with each layer in the three-layer phantom.

Download Full Size | PDF

 

Fig. 6. Measured spectra of three-layer phantom at zero probe-sample separation plotted as a function of wavenumber.

Download Full Size | PDF

The shape and order of the curves in Fig. 7(a) matches the plots in Fig. 7(b) when the FFT curves are aligned with the layer boundaries, which are shown by thick black lines (Fig. 7(b)). For example, at zero probe-sample separation, Fig. 7(b) indicates that the top layer should be preferentially probed by BF0, followed by BF1, BF2 and so on. The relative signal collection efficiencies in the FFT analysis of the top layer are consistent with the pattern observed in the depth profiling plots. Similarly, according to the depth profiling curves (Fig. 7 (b)), the middle layer should be probed best by BF0, followed closely by BF1. The bottom layer would then be expected to follow the pattern with the optimal probing by BF2, then BF3, and BF1. The FFT plots shown in Fig. 7(a) match the predictions for the middle and bottom layers. These results are noteworthy because it provides evidence that the beveled fiber design has the potential for depth-resolved spectroscopic measurements under typical scattering conditions found in precancerous tissue.

 

Fig. 7. Depth resolution measurements of the three-layer phantom. (a) Plot of the amplitude of the primary frequency component for each layer as a function of distance from the probe tip. (b) Depth profiling of C2 configuration of collection fibers using a white substrate with schematic of the three-layer phantom shown above. The boundary between phantom layers are indicated by thick black lines.

Download Full Size | PDF

3.3 In vivo measurements

The 40 degree bevel probe was also used in vivo on oral mucosal tissue of a normal volunteer. Figure 8(a) shows representative measured spectra from the inside of the lower lip and the dorsal tongue of a normal volunteer. Both the lip and tongue spectra display a large range of intensity values for the collection fibers. This can be attributed to differing collection efficiencies and differing photon path lengths sampled for each fiber. Another prominent feature are the dips in the scattering spectra that are consistent with absorption from oxygenated hemoglobin at 540 nm and 576 nm.

The amount of hemoglobin absorption in the scattering spectrum, due to the presence of hemoglobin carrying capillaries in the stroma, is used as a benchmark to ascertain the depth of interrogation of the collection fibers of the beveled probe design. It was expected that the spectra from collection fibers that interrogate more deeply would have larger hemoglobin absorption peaks. One way to estimate the hemoglobin contribution is to measure the amplitude of a hemoglobin absorption peak relative to a reference point that has very small contribution of hemoglobin absorption. Figure 8(c) is a bar graph of the intensity ratio at 576 nm and 610 nm. Normalizations at wavelengths longer than 610 nm display the same trends as in Fig. 8(c). Examination of the beveled fibers show that the hemoglobin absorption contribution increases with distance from the source fiber for both tissue locations, with the BF0 spectrum showing little hemoglobin absorption and the BF5 spectrum showing the largest hemoglobin absorption. These preliminary results indicate that a multifiber beveled probe can obtain depth resolved spectra in vivo.

 

Fig. 8. In vivo measurements of oral mucosa. (a) Representative spectra from inner lower lip and dorsal tongue of a normal volunteer. (b) Spectra normalized to one at 610 nm in order to compare spectral profiles. Hemoglobin absorption appears as dips in the spectra at 540 and 576 nm. (c) Ratio of intensity at 576 nm and 610 nm illustrating changes in relative hemoglobin contribution for the different collection fibers.

Download Full Size | PDF

4. Discussion

As the majority of cancers are epithelial in origin, detection of the earliest precancerous changes in the epithelium, as well as corresponding alterations in the stroma has the potential to greatly impact cancer detection and treatment. Typical flat-tipped spectroscopic probes interrogate a broad range of tissue depths, making it difficult to separate the spectral contributions from the epithelium and stroma. Mathematical algorithms have been developed to aid in the separation of these two signals; however, they often require a priori knowledge of the tissue or assumptions about its optical properties [13, 43, 46–50]. This problem is compounded by precancers that have epithelial thickening and/or tissue keratinization. Use of multiple depth-sensitive fibers can ameliorate this difficulty, where each fiber interrogates a different region beyond the probe tip. The probe design described in this manuscript utilizes a single flat-tipped illumination fiber, combined with multiple beveled collection fibers and a single flat-tipped collection fiber. Based on measurements acquired from a thin white substrate, a 40 degree beveled probe was determined to be the optimal probe for interrogation of epithelial tissue because of its collection efficiency and depth resolution. Figures 2 and 3 demonstrate that the beveled fibers closest to the fiber have the greatest signal intensity and shallowest interrogation depths. The relatively poor signal intensity for beveled fibers farthest from the source fiber indicates that there is an upper limit on source-detector separation for beveled fibers. The presence of this upper limit suggests that, in general, the beveled fiber geometry may not be optimal for deep tissue interrogation. Use of flat-tipped fibers, however, have better collection efficiency at larger depths and may be a better alternative for probing tissue deeply. For example, the FTF in the probe presented here has minimal edge-to-edge separation from the source fiber. At this minimum separation, the peak probing depth is approximately 700 μm from the probe tip with a FWHM of ca. 1760 μm. Increasing the distance of the FTF from the source fiber will correspondingly increase the sampling depth and therefore can be used to complement the shallow-to-intermediate depth information obtained with the beveled fibers.

In a realistic tissue environment, scattering attenuates light propagation and as a consequence the detected signal is a mixture of contributions from different depths. Tissue turbidity can alter the sampling depth of our probe and this was the impetus for performing the experiments with the multilayer bead phantom where the depth dependent scattering properties were known and with in vivo on oral mucosa tissue. Collier et al. [14] has measured the scattering coefficient at 810 nm for cervical tissue to be approximately 22 cm^-1 for normal epithelium, 69 cm^-1 for precancerous epithelium, and 96 cm^-1 for the underlying stroma. Clark et al. [51] obtained similar values for the oral mucosa. These values indicate the photons completely traversing the normal superficial epithelium undergo, on average, one scattering event while photons in precancerous epithelium undergo approximately 3 scattering events or less. The underlying stroma, however, is not optically dilute and will have a much more pronounced effect on the penetration depth of photons. Evaluation of the beveled fiber probe under tissue-like scattering conditions in a three-layer bead phantom showed that the depth profiling curves obtained from the phantom (Fig. 7, a) are very similar to the thin white substrate results (Fig. 7(b)). The width of the depth profiling curves is slightly narrower in the three layer phantom case when compared to the thin white substrate data owing to the turbidity of the phantom. This finding is consistent with Monte Carlo simulations carried out by Pfefer et al. [35] for an obliquely oriented illumination fiber geometry and with data obtained by Schwarz et al. [52] using a ball lens probe in highly scattering media.

Further, the use of multiple optical fibers allows probing of multiple depths within tissue. The in vivo spectra, shown in Fig. 8, are consistent with the expectation that the contribution from hemoglobin absorption from the stromal layer increases for each successive beveled fiber, from BF0 to BF5. The ability to probe multiple depths is important for applications where the epithelial thickness varies. For example, benign conditions in the oral cavity often have hyperkeratinization or hyperplasia. Premalignant and malignant conditions can also have marked epithelial thickening. Within the oral epithelium, the bottom third that includes the basal layer is of interest as the majority of cancers originate and propagate from basal epithelial cells. Variations in epithelial thickness can complicate the specific targeting of this sublayer, the stroma, or of the entire epithelium, and therefore, the ability to interrogate multiple depths becomes increasingly important.

Our results demonstrate that the variability of normal oral epithelial thickness can be observed in vivo. Comparison of the inner lip and tongue spectra indicates epithelial thickening from keratinization in the tongue. This is evident as diminished absorption at the hemoglobin peaks in the tongue spectra obtained using beveled fibers as compared to the corresponding lip spectra in Fig. 8(a), 8(b). Of note is the slope of the curves in the wavelength range 450 – 500 nm. A positive slope can be attributed to strong hemoglobin absorption at 420 nm while the lack of a strong absorption dip, indicated by a negative slope, can be attributed to epithelial thickening from keratinization. For the lip, the signal from all fibers shows a positive slope, except for BF0 signal, which probes most shallowly and, therefore, exhibits very small contribution from hemoglobin. This is consistent with the nonkeratinized structure of inner lip, where the hemoglobin containing capillaries are near the tissue surface [53]. In contrast, the signal from all fibers except the FTF exhibits a negative slope in the tongue spectra, as is expected for keratinized masticatory tissue [53]. Because of the increased epithelial thickness of the tongue, the hemoglobin dips seen in the tongue scattering spectra are less pronounced than in the lip spectra, as shown in Fig. 8(c). It should be noted that the observed differences in hemoglobin absorption are not purely a function of photon propagation in the axial direction through the epithelium and stroma, but also have some contribution from long path length photons that diffuse through the highly scattering stroma into the collection acceptance cones of the beveled fibers. Methods to reduce the contribution of long path length photons include: implementation of a polarization gating scheme [34, 54, 55] or subtraction of adjacent fiber signals for differential path length spectroscopy [27]. The relative contribution of these long path length photons is an active area of interest and is the focus of future work. The probe design presented in this manuscript was not intended to give precise penetration depths, but rather a range of interrogation depths within the superficial tissue that could provide diagnostically useful information. The promising preliminary results from the phantom and mucosa measurements indicate that the use of multiple beveled collection fibers can provide the flexibility necessary to probe a range of interrogation depths, thus, enabling in vivo study of the depth dependent changes that occur with precancer development.

Work by others have shown similar depth sectioning ability using a ball lens probe [32, 52] or more recently, using Monte Carlo simulations to evaluate the combination of a ball lens and beveled fibers [56]. The benefit of the design presented in this manuscript is its manufacturing simplicity and low cost. It has no moving parts and contains few components, permitting the construction of a disposable device. Without fabricated optical components such as a lens, the dimensions of the probe can be potentially miniaturized to diameters less than a millimeter. Such small optical probes can be used within the biopsy channel of traditional endoscopes and surgical needles for use in a broad range of tissues such as breast, colon, bladder, esophagus, and lung and can be used with a variety of optical spectroscopy approaches such as reflectance spectroscopy, fluorescence spectroscopy, and Raman spectroscopy.

5. Conclusions

Depth selective measurements can interrogate the earliest morphological changes that occur with dysplasia. These changes typically start at the basal layer of the epithelium, with coincident alterations occurring in the stroma. A fiber optic probe with beveled collection fibers was evaluated for its optical depth sectioning ability. The results show that a 40 degree bevel provided the optimal depth resolution and signal intensity for study of the depth dependent precancerous epithelial changes. Measurements on a multilayer tissue-simulating phantom and in vivo on the oral mucosa confirm that use of multiple beveled fibers has the capability to provide depth resolved measurements of the optical signatures within epithelium and stroma that has the potential to significantly improve noninvasive detection and monitoring of epithelial precancers.