1. Introduction

Esophageal adenocarcinoma claims 90% of all diagnosed patients within five years [1]. Incidences of the disease have been increasing faster than any other malignancies in the U.S.A. [2]. Up to 44% of presenting patients with advanced disease have no symptoms until the disease have become invasive [3]. At this advanced stage, esophageal resection or chemoradiation becomes the only means of curative therapy with associated morbidity and mortality [4,5] in comparison to less invasive local endoscopic therapies being developed for earlier stage superficial lesions within the mucosa, including mucosal resection and ablative techniques [6–10]. In the currently identified high-risk group such as patients with Barrett’s esophagus, surveillance recommendations have been developed [11]. However, limited visualization of epithelial microstructures during endoscopy restricts the ability to detect early changes of disease progression. Our ultimate goal is to develop a device capable of highly accurate imaging of esophageal disease progression from Barrett’s esophagus to adenocarcinoma, in a heterogeneous cellular population lining the esophageal mucosa.

An array of optical tissue characterization techniques including fluorescence, light scattering, Raman, and Fourier-transform infrared spectroscopy have been explored as diagnostic tools for early detection of esophageal lesions [12–16]. Optical coherence tomography (OCT) has shown ability to provide images of structural components [17,18]. A number of recently developed endomicroscopy systems have also shown promise of increasing sensitivity for disease detection. These include wide-field and confocal endomicroscopy that use visible or white light in conjunction with narrow band imaging or exogenous agents to maximize contrast in the esophagus [19–23]. While these studies have provided promising results, further optimization of this type of technology may expedite its acceptance by the medical community and the health care system.

The objective of this work is to explore the potential of autofluorescence (AF) microscopy under ultra violet (UV) excitation to provide images of human esophagus tissue at the microscopic level. We hypothesize that the superficial photon propagation depth at UV excitation wavelengths will allow high resolution imaging of the epithelial layer without the use of sectioning microscope designs, contrast agents, or tissue preparation. Our preliminary results indicate that microstructure morphology and organization can be imaged in real-time, providing immediate information related to the presence and progress of disease that originates in the epithelium.

2. Materials and Methods

This study was conducted in accordance with an institutional review board (IRB) approved protocol at the University of California, Davis Medical Center. The preliminary results reported in this work represent the findings from an initial population of 30 patients with a history of Barrett’s esophagus.

After obtaining informed consent from patients undergoing routine surveillance for Barrett’s esophagus, standard forceps were used during endoscopy to collect one biopsy specimen from the vicinity of the squamocolumnar junction, and two specimens from the gastroesophageal junction for a total of three biopsy specimens per patient. Each specimen was immediately placed in an individually labeled container with RPMI1640 media (Invitrogen, Carlsbad, CA) and transported to the imaging lab located in the same building where the AF microscopy measurements were taken. All specimens were imaged using ordinary handling. At least three AF images were recorded from specimens that presented high microstructure uniformity (typically observed in normal tissue). Additional images were recorded when tissue images presented altered or changing morphology to best capture these variations. Each image covers a field of view of 670×650 µm² of the surface of the tissue sample. After imaging, each tissue biopsy specimen was immediately placed in 10% formalin for fixation and transferred to pathology for tissue diagnosis. The pathological evaluation was confirmed by at least two expert pathologists and taken as the diagnostic gold standard from which the optical images were categorized. Direct correlation between optical images and histological stain of the same tissue was not possible in this study because the autofluorescence images were acquired along the epithelial/luminal plane, while the histological stained tissue is examined from the plane orthogonal to the surface. Since these biopsy specimens were used for clinical diagnosis, it was not possible to employ an alternate processing technique (slicing from the surface plane).

 

Fig. 1. Schematic depicting the autofluorescence microscopy system. Inset illustrates specimen placement.

Download Full Size | PDF

Each individual biopsy specimen was positioned at the center of a sponge in a standard pathology specimen cassette, which was placed on a xyz translation stage. Ex vivo specimens were imaged without any tissue preparation with a prototype microscope-imaging platform detailed in a previous work [24].

Figure 1 illustrates the experimental arrangement for the imaging of the esophagus tissue specimens used in this study. In brief, a compact diode-pumped solid state laser operating at 266 nm (Intelite, Inc., Minden, NV) was the main light source used to excite the samples and generate the AF images, which were recorded using a liquid nitrogen cooled charge coupled device (Princeton Instruments, Inc., Trenton NJ). Additional compact lasers operating at 355 nm, 405 nm, and a white light source shown in Fig. 1 were used in the preliminary phase of the study that was focused on determining the optimal excitation wavelength.

Figure 2 demonstrates the images of the same specimen under a) 266 nm, b) 355 nm, and c) 408 nm laser excitation along with an image acquired using the same set-up under (d) white light illumination. The contrast in these images was linearly adjusted to qualitatively optimize visualization of the tissue microstructure. The features observed in the white light illumination image (Fig. 2d) were due to artifacts arising from dust on the optical elements. These artifacts are also visible in the AF images under 355 nm and 408 nm excitation, but not under 266 nm excitation. On the other hand, it can be readily appreciated that the images acquired under excitation at 266 nm provided the highest image contrast. For this reason, we focused our attention in the analysis of the experimental results obtained under 266 nm excitation.

The images under 266 nm excitation were acquired using a 5 second exposure time with an approximate dose of 30 mJ/cm². This yielded images with a digitized intensity of about 5000 counts per pixel. It must be noted that high quality images under these excitation conditions were possible with exposure times shorter than one second (or images with digitized intensity of about 500 counts per pixel or higher), but we chose this exposure time in order to optimize image quality. The optical elements used in the microscope were not transmitting UV light, thus the image acquired were based on emission in the visible range. To better quantify the spectral range used for imaging, a 400 nm long pass filter was positioned in front of the CCD. Consequently, the experimental method and results described in this work can be translated for potential future adaptation into commercial endomicroscope systems that are currently capable of imaging only in the visible spectral range.

3. Results

Typical AF images under 266 nm excitation illustrating the observed epithelial morphology associated with various stages of esophageal disease progression are shown below. The images shown were selected as representative of distinct classifications based on the opinion of the expert pathologist. Specifically, Fig. 3(a) shows an image of normal squamous mucosa while Fig. 3(b) that of normal glandular mucosa. AF images of Barrett’s esophagus segments are shown in Fig. 4(a) and Fig. 4(b). High-grade dysplasia is shown in Fig. 5(a), while Fig. 5(b) represents an adenocarcinoma specimen. These images demonstrated visualization of microstructure morphology that could not be attained during a routine endoscopy. Comparing these AF images with histopathology is a crucial step in establishing the foundation necessary for assessing tissue status in vivo. Specifically, we wanted to explore the possibility that information needed for the evaluation of cell organization and morphology can be attained from images such as shown in Figs. 3, 4, and 5.

 

Fig. 2. Image of a 300 µm×270 µm area of a human esophagus columnar epithelium specimen under a) 266 nm, b) 355 nm, c) 408 nm excitation, and d) white light illumination.

Download Full Size | PDF

The image of normal stratified squamous mucosa of the esophagus in Fig. 3(a), showed characteristic tile-like appearance of squamous cells with well demarcated edges at the periphery, probably corresponding to intercellular junctions. The image shown in Fig. 3(b) was from a biopsy specimen collected near the squamocolumnar junction and indicated regular rounded cells with a “honey-comb” appearance of normal glandular mucosa, characteristic of simple columnar epithelium.

Figure 4(a) demonstrates an AF image of a Barrett’s esophagus specimen from non-nodular distal esophageal mucosa. This image exhibits a similar appearance to that of normal glandular epithelium. However, the darker features indicated by arrows may represent goblet cells. This suggested that Barrett’s esophagus could potentially be identified among the glandular distal esophageal epithelium, a capability not yet available to the endoscopist. Another example of diseased tissue that appeared normal during endoscopic examination is shown in Fig. 4(b). This image was taken of a patient biopsy specimen whose pathology revealed dysplasia. The dysplasia was noted to be focal and high-grade rather than low-grade.

 

Fig. 3. Autofluorescence images under 266 nm excitation of a 434 µm×301 µm section of human esophagus of a) squamous mucosa, b) glandular mucosa.

Download Full Size | PDF

 

Fig. 4. Autofluorescence image of human a) Barrett’s esophagus, b) dysplastic glandular epithelium under 266 nm excitation. Arrows designate what could be goblet cells

Download Full Size | PDF

Figure 5(a) illustrates the image of a specimen taken from a patient with high-grade dysplasia and suspicion for adenocarcinoma, reflecting the variation of optical images in relation to the grade of dysplasia and its heterogeneity. Microstructure distortion and change of epithelial morphology was obvious. Early onset of these micro-changes is generally invisible during standard endoscopy and remains undetected until the diseased tissue has reached a symptomatic stage of adenocarcinoma, as illustrated in Fig. 5(b). This biopsy specimen was obtained from the central portion of a visible tumor approximately 5 cm in length. Figs. 5(a) and 5(b) shared similar deviations from normal glandular epithelial cellular structure. Both specimen images demonstrated a loss of the specific cellular patterns that characterized normal epithelium.

High-grade dysplasia was difficult to identify immediately in Fig. 5(a), in part due to the suspicion that this specimen also contained adenocarcinoma. Adenocarcinoma, shown in Fig. 5(b), had a three-dimensional villiform pattern, visibly different from the surface layer of normal glandular cells. Pathological evaluation relies on cross-sectional view of the biopsies to identify features such as nuclear and cellular disorganizations and basement polarity that indicate dysplastic and malignant activity. However, Figs. 3, 4, and 5 demonstrated that there were clear changes in the cellular contour and organization patterns from normal to adenocarcinoma epithelium that could be appreciated with AF microscopy.

 

Fig. 5. Autofluorescence images of human esophagus a) high-grade dysplasia, b) adenocarcinoma.

Download Full Size | PDF

4. Discussion

The imaging approach discussed in this work enables real-time acquisition and display of images depicting the microstructure of the epithelial layer, which can potentially provide diagnostic information. All human esophagus specimens used in this study were unprocessed and unstained, and the images were acquired shortly after the biopsies were obtained in the operating room. Arguably, eliminating the need to prepare the tissue area to be imaged represents a major comparative advantage of this approach. This method, if implemented in vivo, could provide endoscopists with images of cellular morphology to assist in identifying early disease progression of the esophagus. These critical microscopic changes are invisible under currently used standard endoscopes.

AF emission spectra from various tissue types under UV excitation wavelengths shorter than about 300 nm have been previously investigated [25,26]. These results suggested that the emission profiles were similar to that of tryptophan, displaying a peak around 330 nm and a smooth tail extending beyond 400 nm. However, although tryptophan may be the dominant source of AF signal at the tissue level, the emission of some components giving rise to the AF microscopic images presented in this work may be due to various amounts of emission from other tissue fluorophores. We are currently working to address this issue in detail and we plan to report our findings at a later time.

Although the optical technique presented does not have sectioning capability, the results clearly demonstrate that the superficial photon propagation at 266 nm excitation provides imaging of cellular morphology of the epithelial layer. We postulate that the technology presented in our research matches or exceeds the performance of other emerging photonic techniques aimed at providing visualization of tissue microstructure in real-time. For example, Fig. 3(a) presents the superficial layer of squamous cells with visible nuclei that appear to be bright, although nuclei in the subsequent images appear to be dark. This discrepancy may be due to the nature of the stratified epithelia. As the basement layers gradually move toward the lumen of the esophagus, they lose their cuboidal shape for a more squamous surface. The nucleus flattens and undergoes pyknosis (densification) during later stages of maturation. As part of maturation, the nuclei are small and many cells have no nuclei. Imaging cells without nuclei, apoptotic cells, pyknotic nuclei, or the lamina propria surface may produce variable optical results. Nuclei are not clearly visible in the columnar epithelium of Fig. 3(b), probably due to small size and basal location along the basement membrane. Despite this, squamous and glandular mucosa characteristics were easily recognized in the optical images even by a non-expert.

While squamous and columnar epithelia are clearly outlined in Figs. 3(a) and 3(b), a greater challenge would be to identify glandular mucosa from specialized intestinal metaplasia, such as the example in Fig. 4(a). This image contains visible microstructures that are believed to be goblet cells, a defining characteristic of Barrett’s esophagus that is not visible without microscopic capabilities. Dysplastic tissue also cannot be ascertained during routine endoscopic surveillance. Microscopic changes associated with dysplasia were seen in Fig. 4(b). This biopsy specimen was collected from a patient who underwent esophagectomy that showed extensive high grade dysplasia without invasion. The nodular appearance of the AF image most likely corresponds to the polypoid dysplastic epithelium.

The challenge in standard pathology and likely with AF microscopy is to differentiate low grade from high grade dysplasia of the esophagus, particularly in view of the well recognized variation of inter-observer and intra-observer reproducibility. Difficulty of microstructure recognition is compounded by the heterogeneous nature of esophageal adenocarcinoma. Barrett’s esophagus and adenocarcinoma may be present with or without low or high grade dysplasia, increasing the variability of results obtained from the same patient. However, independent of this variability, an optical image using the method discussed in this work is always attainable and may be used to extract diagnostic information when the rules of interpretation to be developed are validated.

Dysplastic epithelium, seen in Figs. 4(b), 5(a), and 5(b), illustrated altered organizational and morphological characteristics of tissue at the microscopic level. These images may be considered as offering a projection of increasingly 3-dimensional cells on the 2-dimensional imaging plane, as opposed to the characteristically flattened 2-dimensional surface of squamous and columnar epithelia in Figs. 3(a), 3(b), and 4(a). The 3-dimensional progression and heterogeneity of tissue morphology associated with disease compounds the difficulty in focusing a single plane by adding an additional variable to the imaging process. Our data is based upon correlating the pathological findings with the AF images of the same specimen. Interpretation of the AF images in normal tissue has relatively higher accuracy and precision in view of the uniformity of cell type, compared to tissues with different cell types (squamocolumnar junction, goblet cells and various grades of dysplasia) which will need more extensive review and correlation of optical images with the pathology. Future work will focus on developing rules of interpretation to enable or assist the utilization of the diagnostic information embedded in the optical images in real-time by the endoscopist.

Obtaining diagnostic information in real-time without contrast agents, sample preparation, time intensive, or prohibitively expensive and complicated instrumentation may be necessary for acceptance of such technology by the medical community and the health care system. This AF method has been implemented for the examination of freshly excised tissue before pathology diagnosis. In this case, UV exposure will not be an issue since low intensity avoids compromise of the tissue sample. While this proof of principle work holds promise that in vivo microscopic AF imaging of tissue is possible via endoscopy, the use of laser excitation, particularly in the UV, will require a design that minimizes exposure and optimizes signal throughput. The AF images presented in this work were acquired under an approximate dose of 30 mJ/cm² but images acquired with one tenth of that exposure were of reasonably high quality (see Materials and Methods section). In addition, the AF images were collected using a 400 nm long pass filter, thus rejecting most of the emission signal under 266 nm excitation. Moreover, the CCD detector was not optimized for highest sensitivity in this spectral region (400–500 nm). Therefore, we believe that a system designed with similar optical parameters but optimized for signal detection should easily meet the maximum permissible exposure (MPE) designated ANSI standard for exposure to UV wavelengths between 180 nm–302 nm of 3 mJ/cm² [27]. Ultimately, in vivo AF microscopy may provide information for disease diagnosis that may help detect esophageal disease at an earlier stage. We postulate that the same technique may be suitable for application in other tissue systems. In the case of the current gold standard, stained tissue sections are imaged and interpreted using standard methods that have been developed to assist diagnosis such as the Vienna classification system [28]. The Vienna classification is a standard used with the exception of category 4.2 noninvasive carcinoma (carcinoma in situ), which is not used. Diagnosis of disease with AF microscopy might require the development of analogous rules.