1. Introduction

A significant limitation to traditional white light endoscopy is the inability to image cellular level epithelial morphology. Emerging techniques that address this problem include confocal fluorescence endomicroscopy, which provides in vivo information using intravenous fluorescein [1,2]. The combination of targeted peptide probes for complementary functional confocal data has also been explored [3]. In addition, wide-field endomicroscopy prototype systems include the use of contrast agents such as acriflavine hydrochloride [4], as well as quantum dots and gold nanoparticles [5]. The use of contrast agents is an additional step that may increase the cost and time budget of the procedure and represents an additional risk to the patient. Consequently, the development of imaging techniques that do not require the use of contrast agents may be desirable. Such techniques may rely on intrinsic tissue chromophores that can be excited via linear [6] or non-linear microscopy techniques [7,8]. The latter method offers a sectioning capability needed to image a specific layer of the tissue and reject out of focus signal. It was shown very recently that the short penetration depth of ultraviolet (UV) excitation gives rise to autofluorescence from only the superficial tissue layer [9]. This in turn allows for imaging of the superficial tissue layer using wide-field microscopy approaches without the need to stain the tissue with contrast agents, employ optical sectioning techniques that reject most of the signal produced by the excitation light, or mandate time intensive tissue preparation.

Several endogenous fluorophores absorb in the UV spectral region and contribute to tissue autofluorescence (AF) emission in the visible spectrum including tryptophan, elastin, collagen, nicotinamide adenine dinucleotide (reduced form NADH), and flavin adenine dinucleotide (FAD) in many different organs such as the head-neck [10] and breast [11]. Investigation of AF in gastrointestinal tissues has been performed at excitation wavelengths longer than 330 nm [12,13]. Tryptophan has been shown to dominate the emission profile under UV excitation shorter than 310 nm [14,15].

The goal of this work is to understand the origin at the microscopic level and spectral characteristics of the AF from human esophagus tissue under 266 nm and 355 nm excitation. The choice of these excitation wavelengths was based on recent results that demonstrated microscopic imaging of unprocessed esophageal mucosa using wide-field microscopy to capture the AF produced with excitation at these wavelengths [9]. As the preliminary results suggest that this approach may enable in vivo pathological assessment with no tissue preparation, understanding the exact mechanism giving rise to image formation is critical for optimizing instrumentation and methodology. The experiments employ AF point spectroscopy and microscopic narrow-band imaging (NBI) to investigate ex vivo normal squamous and columnar mucosa of fresh, unprocessed human esophagus specimens.

2. Materials and methods

Fresh human tissue biopsy specimens were collected from consented patients with a history of Barrett’s esophagus (BE) undergoing routine surveillance. Standard forceps were used during endoscopy to collect one biopsy specimen from the vicinity of the squamocolumnar junction (Z-line), and one biopsy specimen from the gastroesophageal (GE) junction for a total of two biopsy specimens per patient. The protocol was approved by the University of California, Davis Medical Center Institutional Review Board.

2.1 Point spectroscopy

Point spectroscopy experiments were performed with an initial population of four patients collecting two biopsy specimens per patient, for a total of eight tissue samples. Each unprocessed esophagus tissue biopsy specimen was individually placed between two quartz slides to acquire the AF spectra using two excitation lasers operating at 266 nm and 355 nm (Intelite, Inc., Minden, NV) having an average power of about 1 mW. The lasers were aligned and coupled to a UV compatible fiber probe used to collect and transmit the emission from the target area to a spectrometer containing a single excitation delivery fiber (400 µm diameter) surrounded by 6 collection fibers (200 µm diameter). External shutters allowed manual control of laser source exposure. Spectra from four different locations were taken from each tissue specimen under each excitation wavelength and averaged to produce one mean spectrum per specimen. The emission arising from 266 nm excitation was passed through a 280 nm long pass filter, while that under the 355 nm laser was passed through a 385 nm long pass filter in order to reject the excitation light. A flip mirror was aligned to steer the excitation beam along an alternate path for AF NBI image collection described in the next section. Tungsten and deuterium lamps (Oriel Instruments, Stratford, CT) were used to generate calibration curves to correct for the spectral response of the system used to record the AF spectra.

2.2 Narrow-band AF

Narrow-band AF multispectral microscopic images were acquired from an initial population of thirteen patients. Two biopsy specimens per patient were collected for a total of twenty six tissue samples. A minimum of three AF images were recorded from each specimen using each narrow-band filter. A description of our AF microscopy approach and experimental system has been previously described [9]. Baseline AF images from each specimen were acquired under 266 nm and 355 nm excitation using a 400 nm long pass filter under 5 second exposure. Subsequently, each site was imaged using the set of narrow-band filters with a peak transmission wavelength centered from 450 nm to 600 nm in 50 nm increments and a bandwidth of ± 20 nm at full width half maximum (FWHM). Each filter was manually interchanged to collect corresponding spectral images under 30 second exposure time. Images were processed using WinView software (Princeton Instruments, Trenton, NJ). Ratio images were obtained via pixel-by pixel division of NBI results recorded at different spectral bands to highlight the contribution and localization of different fluorophores [16]. Tissue samples were immediately placed in formalin after completion of the experiments and returned to the grossing lab for histopathology diagnosis.

3. Experimental results

3.1 Point spectroscopy

The spectroscopy experiments show characteristic emission bands that can be assigned to three main tissue fluorophores. Under 266 nm excitation, emission band 1 is observed as shown in Fig. 1 centered between 320 nm – 350 nm. This emission band is characteristic of tryptophan emission. The tail of the emission at longer wavelengths does not contain features that can be identified as the contribution from additional tissue fluorophores. This may indicate that tryptophan is the main contributor in the emission spectrum under 266 nm excitation through the entire spectral range of our measurement. However, arrows 1 indicate a region of visible spectral difference in the tail of the emission spectra obtained from different specimens that may arise from other tissue fluorophores, or can be an artifact arising from variation in blood concentration within each specimen (via re-absorption of the emission by blood cells).

 

Fig. 1 Normalized mean autofluorescence spectra of esophagus mucosal biopsy specimens under 266 nm excitation and 355 nm excitation.

Download Full Size | PDF

Under 355 nm excitation, emission bands 2 and 3 are observed between 400 nm – 430 nm and 440 nm – 470 nm that can be assigned to emission from collagen and NADH, respectively. There are no additional features in the measured spectra under 355 nm excitation that can be identified as contribution from other tissue fluorophores, such as FAD or lipofuscin. However, variability in the relative strength of emission bands 2 and 3 was observed. This may be due to blood re-absorption of the emission (as discussed above) and/or on the tissue pathology. Since this is not the focus of this investigation, we will not expand the discussion on this point. Arrow 2 indicates the spectrum of the biopsy specimen further examined with NBI and shown in Fig. 2 .

 

Fig. 2 142 µm x 136 µm raw images of a single ex vivo human esophagus columnar mucosa biopsy specimen under (a) 355 nm excitation and (b) 266 nm excitation with a 400 nm long pass filter and (c) the ratio image of (b) divided by (a). NBI under 266 nm excitation using the (d) 450 nm, (e) 550 nm and, (f) 600 nm filters. Ratio of spectral images (g) 450 nm/400 lp, (h) 550 nm/400 lp and, (i) 600 nm/400 lp.

Download Full Size | PDF

3.2 Narrow-band AF imaging

Figure 2 shows a series of spectral and spectral ratio images obtained from the same location of a biopsy specimen collected from an overlapping Z-line and GE junction at 40 cm. These images represent a typical pattern consistently observed when imaging normal (columnar or squamous) esophageal mucosa of different patients. The gold standard pathology for this specimen was columnar mucosa with mild chronic inflammation.

Figures 2a and 2b are the baseline AF images (referred to as $I_{400lp}^{355}$ and $I_{400lp}^{266}$ ) acquired under 355 nm and 266 nm excitation respectively, using a 400 nm long pass (lp) filter. These two images are remarkably different, with the image under 355 nm excitation (2a) providing very little contrast while the image under 266 nm excitation (2b) providing a clear visualization of the columnar mucosa seen as the characteristic honeycomb pattern. Based on the spectroscopy results, the main difference between these images is that under 266 nm excitation (2b), the AF image is dominated by the tryptophan emission which is absent under 355 nm excitation (2a). Based on the broad absorption spectrum of the fluorophores excited under 355 nm illumination, we can assume that these fluorophores also contribute to the emission under 266 nm excitation. The ratio image $I_{400lp}^{266}$ / $I_{400lp}^{355}$ can remove or at least reduce the contribution of the other fluorophores and thus provide a mostly tryptophan based image. This ratio image is shown in Fig. 2c exhibiting a moderately enhanced contrast in the visualization of the honeycomb pattern.

Figures 2d-2f are raw AF images under 266 nm excitation using narrow-band filters centered at 450 ± 20nm (2d), 550 ± 20nm (2e), and 600 ± 20nm (2f) referred to as $I_{450nb}^{266}$ , $I_{550nb}^{266}$ , $I_{600nb}^{266}$ , respectively. The spectral range used to acquire image $I_{450nb}^{266}$ (2d) is centered at the emission peak of NADH and therefore, the relative contribution of NADH emission in this image should be higher compared to any other narrow-band images. Similarly, flavoproteins and/or lipo-pigments AF should provide the highest relative contribution in the $I_{550nb}^{266}$ image (2e). Finally, the $I_{600nb}^{266}$ image shown in Fig. 2f was recorded at a spectral range that is out of resonance with the emission peak of all major fluorophores contributing to tissue AF, but may still be a significant contribution from lipo-pigments as their emission spectrum is broader than that of the flavoproteins. Despite the careful selection of these narrow band images aimed at highlighting the contribution of additional tissue fluorophores in the formation of the AF image under 266 nm excitation, these images present very minimal difference in image contrast suggesting that tryptophan emission dominates the autofluorescence signal used for image formation within the entire spectral range. This also indicates that tryptophan is the key contributor giving rise to the observed contrast between the cytoplasm and membrane regions, allowing for visualization of the cellular morphology.

As mentioned above, the $I_{450nb}^{266}$ image should contain the highest relative contribution of NADH emission compared to any other spectral image. We hypothesized that a ratio image obtained by dividing the $I_{450nb}^{266}$ image by the $I_{400lp}^{266}$ (which contains the emission of all fluorophores contributing to the detected AF under 266 nm excitation) would provide an image that highlights the localization of NADH at the microscopic level. Similarly, the ratio image obtained by dividing the $I_{550nb}^{266}$ image by the $I_{400lp}^{266}$ may provide visualization of the localization of flavoproteins and/or lipo-pigments. These images are shown in Figs. 2g and 2h, respectively. The ratio image obtained from the division of the $I_{600nb}^{266}$ image by the $I_{400lp}^{266}$ is also shown in Fig. 2i. These three ratio images show the cytoplasm region with higher intensity indicating the localization of the corresponding fluorophores within this region.

A careful examination of these ratio images indicates that the regions of higher intensity are not generally overlapping. To better quantify this effect, the digitized intensity profile along a small section of the images of the specimen shown in Fig. 2 is shown in Fig. 3 . These profiles represent the normalized average intensity over a zone of the imaged area that is 1.2 µm thick spanning through 4 cells. The results are represented as percent change from the average intensity to enable a quantitative assessment of the differences in intensity within each cell as well as facilitate assessment of the image contrast. The profile shown in Fig. 3a shows the intensity of the $I_{400lp}^{266}$ image in this section of the specimen corresponding to the image shown in Fig. 2b. Similarly, The profiles shown in Figs. 3b, 3c and, 3d show the intensity along the same section of the specimen of the $I_{450nb}^{266}$ / $I_{400lp}^{266}$ , $I_{550nb}^{266}$ / $I_{400lp}^{266}$ and, $I_{600nb}^{266}$ / $I_{400lp}^{266}$ ratio images, respectively, corresponding to the image shown in Figs. 2g, 2h and, 2i. The peaks in the profile of Fig. 3a correspond to the location of membranes. There is more than 30% difference in intensity between the emission of the cytoplasm and that of the membrane in the recorded images allowing for a clear visualization of the microstructure of human esophagus columnar mucosa. Figures 3b-3d demonstrate the variation in the intensity in the cytoplasm region in the ratio images obtained from different narrow band spectral windows, which supports the hypothesis that these images arise from different fluorophores.

 

Fig. 3 Digitized intensity profile along the same 1.2 µm zone spanning through 4 cells from a section of the specimen shown in Fig. 2 corresponding to the images obtained (a) under 266 nm excitation with a 400 nm long pass filter, and the spectral ratio images (b) $I_{450nb}^{266}$ / $I_{400lp}^{266}$ , (c) $I_{550nb}^{266}$ / $I_{400lp}^{266}$ , and d) $I_{600nb}^{266}$ / $I_{400lp}^{266}$ .

Download Full Size | PDF

The nuclei of the esophagus columnar epithelium are not visible using this technique because they are located deep below the surface that is outside the imaging depth of this technique. Columnar epithelial cells appear as tall columns with elongated nuclei located towards the basal surface. This is not the case for squamous epithelium. Stratified squamous epithelial cells of the esophagus appear as scale-like tiles that have flattened surface layers. Nuclei are located very close to the surface (lumen) and progressively condense and flatten during maturation. An AF image under 266 nm excitation of a 340 µm x 180 µm region of squamous mucosa biopsy specimen is shown in Fig. 4a . In accordance with the observations established in the study of columnar mucosa (see Fig. 2), the enhanced emission of the membrane and/or intercellular junctions leads to visualization of a tile-like appearance of polygonal cells with well demarcated edges at the periphery, characteristic of this type of tissue. In addition, the circular structures observed within the cytoplasm region of each cell are believed to be the nuclei of the cells. It is therefore possible to obtain information about the nucleus to cytoplasm volume ratio, which is a critical characteristic change directly related to progression of disease such as cancer.

 

Fig. 4 (a) 340 µm x 180 µm raw image of a single ex vivo human esophagus squamous mucosa biopsy specimen under 266 nm excitation with a 400 nm long pass filter. (b) Digitized intensity profile along a 1.5 µm zone spanning through 3 cells from a section of the specimen shown in Fig. 2.

Download Full Size | PDF

Figure 4b shows the digitized intensity profile along a small section of the image of the specimen shown in Fig. 4a over a 1.5 µm wide zone of the of the imaged area of the sample spanning through 3 cells. This profile was obtained using the same method described above to obtain the profiles shown in Fig. 3. The peaks in the profile denoted as “N” and “M” represent the location of nucleus and membrane, respectively. Comparison of the profiles shown in Figs. 3a and 4b indicate that the contrast of the membrane is increased in the case of columnar mucosa (Fig. 3a) compared to that of squamous epithelium (Fig. 4b). This can be attributed to the increased depth of the columnar cells.

4. Discussion

The results suggest that visualization of the epithelial morphology based on its native fluorescence under UV excitation using wide-field microscopy is based on two main mechanisms. The first mechanism is associated with the property that UV light only superficially penetrates epithelial tissue, on the order of 100 µm or less. As a result, the fluorescence signal produced in this superficial tissue layer can be contained within thecomparable thickness of the image plane of the microscope providing high contrast images without using an optical sectioning technique (such as confocal microscopy) that generally causes a large portion of the generated signal to be rejected. That is because the out of focus (background) signal is sufficiently reduced to allow the formation of high contrast images of tissue microstructures using the AF of the tissue cells and intracellular components. It must be noted that this mechanism allows image acquisition based on the emission of all native tissue fluorophores (as they can all be excited with UV light) and it is independent of the emission wavelength. It is therefore possible to acquire images that probe the various optically active analytes to obtain not only structural information but also functional information. In addition, images based on the emission of contrast agents can be attained and combined with those of native fluorophores (if their emission is outside the spectral range of native fluorophores, such as at wavelengths longer than about 750 nm) to provide molecular (or other types) of targeting information.

The second mechanism leading to the acquisition of images that delineates the different compartments of cells is that there is sufficient variability in the concentration of chromophores contained within these compartments. The experimental results presented in this work indicate that tryptophan is the native tissue fluorophore providing the best image contrast using this imaging approach enabling visualization of the microstructure and organization of the superficial layer in a similar way to that provided by H&E staining.

The ratio images shown in Fig. 2 do not provide any additional information to improve the visualization of normal human esophagus columnar mucosa. However, the target-like feature located in the lower right corner of all images of this specimen appears with increased contrast in the ratio images. We believe this feature may represent a villi crypt, which was observed in multiple locations of this specimen as well as in other specimens. It is therefore possible that spectral ratio imaging may be useful in providing additional diagnostic information that can enhance the ability to evaluate tissue pathology.

Understanding the exact mechanism for image formation was essential for establishing the optical criteria necessary for differentiation between Barrett’s esophagus (BE) and grades of dysplasia, from low grade through esophageal adenocarcinoma, as well as to develop the designing criteria for implementation in vivo using endomicroscopy. Our work in these two areas is in progress and results will be reported in the near future.

Incorporation of NBI may be more difficult to implement in a clinical in vivo setting, but has the potential to visualize epithelial morphology for early disease detection where biochemical changes precede morphological changes. In addition, it can provide functional information using intrinsic tissue chromophores while molecular targeting information can be added with the use of contrast agents. Image multiplexing is possible using this technique and, depending on instrumentation design, it can be implemented using parallel (simultaneous) multi-image acquisition via image splitting to different spectral bands. This can be followed by reconstruction of the image to its different principal components to delineate the structural, functional, and molecular/targeting information.