- Email: [email protected]
Comparative optical coherence tomography imaging of human esophagus: How accurate is localization of the muscularis mucosae?
Comparative optical coherence tomography imaging of human esophagus: How accurate is localization of the muscularis mucosae? Inci Çilesiz, PhD, Paul Fockens, MD, PhD, Raphaela Kerindongo, Dirk Faber, MS, Guido Tytgat, MD, PhD, Fiebo ten Kate, MD, PhD, Ton van Leeuwen, PhD Istanbul, Turkey, and Amsterdam, The Netherlands
Background: Early diagnosis of esophageal cancer limited to the mucosa allows local endoscopic treatment and thereby improves prognosis. Optical coherence tomography images of normal human esophageal tissue obtained with 2 systems with light sources that provide different wavelengths (800 nm and 1275 nm) were compared with histology to determine which wavelength is best suited for detailed optical coherence tomography imaging of the esophageal wall, and to precisely localize the muscularis mucosae. Methods: Within 1 hour of surgical resection, an esophageal specimen was cleaned of excess blood with saline solution and soaked in formalin for a minimum of 48 hours. After optical coherence tomography imaging, the specimen was prepared for routine histologic assessment. To precisely localize the different layers of the esophageal wall on an optical coherence tomography image, well-defined structures within the esophageal wall were sought. Results: The 1275 nm system with 12 µm resolution was superior in terms of imaging depth. As compared with histology, the 4 µm resolution of the 800 nm system made fine detail more visible. With minimal experience, the muscularis mucosae could be recognized with either system as a hyporeflective layer with a diameter of around 180 µm. Conclusions: Based on appearance and location of morphologic landmarks, layers of normal esophageal wall, specifically, the location and extent of the muscularis mucosae, could be recognized by using both the 800 nm and 1275 nm optical coherence tomography system. Although different conditions may be operative in vivo, the present ex vivo study further verifies by precise interpretation that optical coherence tomography provides precise images of the esophageal wall. (Gastrointest Endosc 2002;56:852-7.)
The incidence of adenocarcinoma associated with Barrett’s esophagus is rising rapidly in the developed world.1,2 Accurate diagnosis plus intervention before cancer has invaded beyond the muscularis mucosae will hopefully improve prognosis for patients with this disease.3,4 Because existing methods are adequate for removal of early stage lesions in a minimally invasive manner, accurate in vivo delineation of dysplastic lesions and early stage canReceived December 28, 2001. For revision May 8, 2002. Accepted July 29, 2002. Current affiliations: Biomedical Engineering Program, Electronics and Communication Engineering Department, Istanbul Technical University, Istanbul, Turkey, and Department of Gastroenterology, Lasercentrum, and Department of Pathology, Academisch Medisch Centrum, Amsterdam, The Netherlands. Part of the research program of FOM [Stichting voor Fundamenteel Onderzoek der Materie, financially supported by NWO, the Nederlandse Organisatie voor Wetenschappelijk Onderzoek](IÇ, DF), and Philips Research (IÇ). This work was also funded by ZON [Zorg Onderzoek Nederland](RK). Reprint requests: Ton van Leeuwen, PhD, Lasercentrum, Meibergdreef 9, 1105AZ Amsterdam, The Netherlands. Copyright © 2002 by the American Society for Gastrointestinal Endoscopy 0016-5107/2002/$35.00 + 0 37/1/129606 doi:10.1067/mge.2002.129606 852
GASTROINTESTINAL ENDOSCOPY
cers is essential for selection of optimal treatment and minimizing morbidity. The current diagnostic method of choice, endoscopy with white light illumination, offers remarkable capabilities including noninvasive visual access to inner organs and direct tissue sampling. Nevertheless, visualization of only the luminal aspect of the mucosa under relatively low magnification places microscopic clues to underlying pathophysiology beyond the reach of the endoscopist.5 In general, few endoscopic features permit differentiation of nondysplastic and dysplastic mucosa. Moreover, early stage invasive cancer may not be visible endoscopically and may not be detected without biopsies, a method hampered by a substantial sampling error.3,6 A recent study involving histopathological diagnoses based on biopsy specimens from 1068 patients who underwent endoscopy because of gastroesophageal reflux disease found that 92% of cases of dysplasia and 33% of carcinomas in patients with Barrett’s esophagus were undetected at endoscopy by experienced endoscopists.7 EUS, even at high frequency (30 MHz), provides cross-sectional images with relatively low spatial resolution (100 µm at best).5,8,9 In contrast, optical VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
I Çilesiz, P Fockens, R Kerindongo, et al.
Figure 1. Simplified diagram of the fiberoptic Michelson interferometer used for OCT imaging. RSOD, Rapid scanning optical delay line.
coherence tomography (OCT), a newly emerging high resolution imaging technique for analysis of subsurface tissue structures to a maximum depth of 2 mm, is an excellent candidate method for determining noninvasively and in vivo the location, extent, and progression of normal and altered structures, including dysplasia, by guiding and supplementing endoscopy. Preliminary investigations have demonstrated the feasibility and superiority of OCT imaging systems with 1300 nm light sources for visualization of tissue layers.5,10-14 Analogous to B-mode US imaging, OCT imaging provides 2-dimensional cross-sectional morphologic images with spatial resolution approaching cellular level (i.e., ≤10 µm laterally and axially), leading to real-time visualization of tissue substructure that is nearly equivalent to standard microscopy of histologic sections. The characteristics of photon propagation in tissue, and consequently the attributes of OCT images, are primarily governed by the local optical properties of tissue: (1) absorption, (2) scattering coefficients, and (3) degree of scattering anisotropy of different layers of tissue at varying wavelengths of the optical spectrum.15 Theoretically, any morphologic formation unlike the normal tissue substructure will cause a local change in optical properties, particularly scattering coefficient and anisotropy, and will thus modify the optical behavior of tissue, which will in turn be detectable by OCT. Hence, OCT imaging has revolutionary potential to greatly minimize the sampling error associated with standard endoscopy and random biopsies and to open the possibility of early endoscopic diagnosis of dysplasia in situ. VOLUME 56, NO. 6, 2002
This study investigated the feasibility of using 2 laboratory OCT systems, based on different light sources (Ti:Sapphire laser, around 800 nm; a super luminescent diode, around 1275 nm), to characterize esophageal tissue layers with special emphasis on localization of the muscularis mucosae (MM). The study rationale was that cancer invasion beyond the MM may preclude local endoscopic treatment and thus confers a worse prognosis for patients. MATERIALS AND METHODS Tissue specimen An area of normal-appearing proximal esophagus was removed from the resection specimen from a patient who underwent operation for gastric cardia carcinoma. The tissue specimen was obtained within 1 hour after resection. Excess blood was removed by washing with saline solution and the muscularis propria was stripped off. The specimen was then stretched and pinned on clear agar gel in a Petri dish. The prepared specimen was soaked in formalin for a minimum of 48 hours for optimal preservation of tissue morphology. Before OCT imaging the specimen was submerged in normal saline solution. Imaging sites were marked on a drawing of the tissue sample (tissue map) for OCT and histology sections. The estimated correspondence among imaging sites and histology sections was ± 1 mm. After OCT imaging with both systems, imaging sites were prepared and stained with hematoxylin & eosin. Finally, the specific layers in histologic sections viewed under a standard light microscope were compared and contrasted with OCT images. OCT imaging Comparable OCT images of human esophageal tissue were acquired with one OCT system based on a commercially available femtosecond Ti:Sapphire laser (TSL) at GASTROINTESTINAL ENDOSCOPY
853
I Çilesiz, P Fockens, R Kerindongo, et al.
OCT imaging of human esophagus, localization of muscularis mucosae
Figure 2. OCT images and comparative histologic section of normal esophageal wall with an embedded ellipsoid structure. A, OCT image with TSL-based system. B, Photomicrograph of comparative histologic section showing lymphoid follicle (lf) embedded between lp and mm (H&E, orig. mag. ×40). C, OCT image with SLD-based system. ep, Epithelium; lp, lamina propria; mm, muscularis mucosae; sm, submucosa. 800 nm and a second OCT system based on a super luminescent diode (SLD) at 1275 nm. Technical specifications for the 2 systems were as follows: TSL system, optical bandwidth, 90 nm full width at half maximum (FWHM); axial resolution, less than 4 µm in tissue; optical power incident on sample/spot size, 4 mW/5 µm diameter; dynamic range (detection sensitivity), –115 dB. SLD system, optical bandwidth, 50 nm FWHM; axial resolution, approximately 12 µm in tissue; optical power incident on sample/spot size, 0.5 mW/20 µm diameter; dynamic range (detection sensitivity): –105 dB. Figure 1 is a simplified diagram of the optical fiber-based OCT systems. The probing/scanning depth was gated by low coherence interferometry with the tissue sample placed in the sample arm of a Michelson interferometer and a double pass rapid scanning optical delay (DP-RSOD) line in the reference arm.16 The DP-RSOD line was driven by a round-off triangular waveform at 50 Hz, allowing a scan depth of 1.5 mm. Lateral scanning was accomplished by translation of either the sample (TSL-based system) or the sample beam (SLD-based system). Analogous to US imaging, a two-dimensional OCT image (brightness or Bscan) was composed of a series of adjacent axial depth scans (amplitude or A-scans). In experiments with the TSL-based system, each 4 mm long B-scan consisted of 400 A-scans separated by 10 µm. Each 3.4 mm long B-scan in the SLD-based system consisted of 500 A-scans separated by 6.8 µm. Detected light intensity was optimized for image contrast with 256 gray levels. To facilitate the distinction between different layers of the esophageal wall (epithelium [EP], lamina propria [LP], MM, submucosa [SM]), and to precisely localize these layers on an OCT image, well-defined and easily identifiable features were sought that appeared repeatedly in a certain location within the esophageal wall. Imaging sites that had such landmark features were reimaged with 5 to 10 B-scans separated by 100 to 500 µm. Such sites were marked on the tissue map to guide histologic sectioning. Guided histologic sections were then studied to facilitate resolution of the location of the land854
GASTROINTESTINAL ENDOSCOPY
mark features within the esophageal wall, and to more accurately interpret OCT appearance of the layers adjacent to such landmark features.
RESULTS OCT was first performed on a total of 27 samples taken from the esophageal specimen with the TSLbased system. The layered structure of the esophageal wall was clearly identified in all images, the layers being distinguished by the relative differences in the intensity of their backscattered reflection. Based on a comprehensive review of all of these images, several well-defined dark ellipsoid structures embedded within the esophageal wall were discovered. After a second review of all OCT images and the acquisition of repeated OCT images from 8 of these imaging sites, a total of 4 imaging sites believed to contain the ellipsoid structures were selected for imaging with the SLD-based system. These ellipsoid structures were not visible to the naked eye with the specimen soaked in formalin or saline solution. However, while the samples were dried and prepared for histologic processing, clusters of the ellipsoid structures were identified over an area of approximately 2 cm2 just underneath the esophageal epithelium by using a low-power magnifying glass. Histologic sections of the well-defined ellipsoid structures revealed the typical appearance of lymphoid follicles. Examples of histologic and comparative OCT images obtained from such a site are shown in Figure 2. The follicles appeared darker (less reflective) than the surrounding structures in all OCT images and were 300 to 600 µm wide, 250 to 300 µm thick, and up to 500 µm long. Their repeated appearance between LP and MM facilitated precise localization of LP and MM and identification of VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
I Çilesiz, P Fockens, R Kerindongo, et al.
Figure 3. OCT images and comparative histologic section of normal esophageal wall: A, OCT image with TSL-based system. B, Photomicrograph of comparative histologic section (H&E, orig. mag. ×40). C, OCT image with SLD-based system. ep, Epithelium; lp, lamina propria; mm, muscularis mucosa; sm, submucosa.
the first 3 layers of the esophageal wall (EP, LP, MM) in all imaged sections. In the positive (commonly referred to as grayscale) OCT image display modality used for both systems, bright white corresponds to highly reflective whereas dark gray or black corresponds to less reflective. An additional histologic section and OCT images obtained with both OCT systems are shown in Figure 3. In OCT images of normal esophageal wall, desquamating cells appeared as a thin bright (highly reflective) line. The EP appeared darker (less reflective), and the LP slightly brighter (more reflective). The interface between EP and LP was relatively easier to recognize on OCT images obtained with the SLDbased system. MM appeared darker (less reflective) than LP and also “wavy” on images acquired with either system. However, the wavy pattern was easily recognized on images obtained with the TSL-based system, but not as obvious in those with the SLDbased system (Fig. 3A and C). Nevertheless, the boundaries of the MM could be identified by “lines” running horizontally above and under the MM in images acquired with both systems. These lines indicated change of structure and scattering properties, that is, interfaces with LP and SM. From a total of 77 B-scans with and without follicles present, the MM was identified as a 180 ± 22 µm thick hyporeflective layer. Boundaries of the MM could additionally be identified by the brighter (more reflective) appearance of the SM in OCT images obtained with the SLD-based system; whereas for images acquired with the TSL-based system, “visibility” was greatly reduced beyond the MM and the SM appeared dark. Loss of visibility at depths greater than 500 to 600 µm in the tissue (Fig. 3A) may be caused by relatively shallower penetration of light and higher scattering properties at 800 nm compared with 1275 nm. In VOLUME 56, NO. 6, 2002
general, more detail could be observed in images acquired with the TSL-based system, as anticipated based on the 3-fold greater resolution of that system. DISCUSSION To our knowledge, this is the first comparison study of histology versus OCT images from the normal esophageal wall with 2 different systems with 2 different wavelengths, in which distinct, well-defined morphologic markers such as lymphoid follicles are shown. The LP throughout the digestive tract contains cells belonging to the immune system. Lymphoid cells are commonly arranged as large follicles partly within the mucosa and partly within the SM, splitting the MM.17 In contrast, the esophageal LP contains scattered lymphocytes. Nevertheless, aggregations of lymphoid cells forming small follicles are common in the esophageal SM near the squamous/columnar epithelial junction and occasionally in upper parts of normal esophageal SM.17 Consequently, the frequent presence of lymphoid follicles in the upper histologic layers of the esophageal wall away from the squamous/columnar epithelial junction was considered a manifestation of the inflammation/immune response to chronic gastroesophageal reflux. Although only one esophagus specimen was included in this study, OCT images were examined from 3 different pathologic specimens from 3 different patients and similar follicles were observed. The time required for preparation was too long for 2 of these specimens, so that tissue decomposition could influence the optical properties of the different tissue layers; thus, these 2 specimens were not included in the study. Note that in histologic sections, periesophageal mucous glands may appear similar to follicles (Fig. 2). The different layers of normal esophageal wall could be precisely identified with both the 800 nm– GASTROINTESTINAL ENDOSCOPY
855
I Çilesiz, P Fockens, R Kerindongo, et al.
and 1275 nm–based systems by comparison with corresponding histologic sections. These results indicate that the SLD-based OCT system with 12 µm axial resolution was superior in terms of penetration and visibility of deeper layers compared with the TSL-based system. Based on analysis of the histologic findings, the 4 µm resolution of the TSLbased system made fine detail and structures clearly recognizable. Although a better resolution with the SLD-based system would improve visualization of detail, with minimal experience the interface between LP and MM could be recognized by using either system. In addition to increased imaging depth, an SLD-based system with slightly less resolution may be preferable for clinical esophageal OCT imaging because it is portable, less costly, and easier to maintain when compared with a bulkier, more expensive, TSL-based system. Other groups have demonstrated the feasibility of endoscopic high-resolution OCT imaging in the GI tract.5,12-14 However, these investigators have compared histologic sections with OCT images acquired with 1300 nm systems of varying optical bandwidths. Although separation between SM and muscularis propria as a result of saline solution injection has been captured on an OCT image,13 no morphologically well defined structures were sought as a means to interpret OCT images and localize the MM. In general, OCT images and histologic sections from cadaveric or autopsy specimens have been used to study the layered structure of the human esophageal wall and to interpret OCT images obtained in vivo.18 In the first report of GI OCT imaging, autopsy specimens were used to study esophageal structures.19 By using negative (i.e., black on white) OCT imaging, these investigators stated that MM appeared brighter than the mucosa and that gaps were observable between the MM and SM. In a positive OCT image display, also referred to as grayscale (white-on-black), “brighter” or “whiter” signifies “more reflective,” whereas “darker” or “grayer” signifies “less reflective.” In contrast, in a negative OCT display, also referred to as inverse grayscale (black-on-white), “brighter” signifies “less reflective,” whereas “darker” signifies “more reflective.” To avoid confusion, an independent image display terminology was chosen, and the terms “more reflective” and “less reflective” were used. The boundaries of the MM as observed in the present study appeared more reflective. However, the MM was in general less reflective than the LP and appeared “wavy,” an observation not previously reported. The explanation for this observation is believed to be two-fold: the relatively higher resolution (4 µm vs. 10-20 µm in earlier studies) made fine 856
GASTROINTESTINAL ENDOSCOPY
OCT imaging of human esophagus, localization of muscularis mucosae
detail more visible, and by stretching and pinning the sample on clear agar gel, the EP appeared to be 100 to 150 µm thick, which is much thinner than the 500 to 800 µm previously reported.17 Stretching the sample also flattened the irregular lower border of the EP. Therefore, back-scattered photons of light coming from deeper layers of the esophageal wall might have had a better chance to propagate toward the lumen without being scattered by irregular borders. The flattening of EP may have thus “improved” detection sensitivity for signals coming from deeper layers, thereby providing greater structural detail. Gaps were frequently observed between and within the layers of the esophageal wall. These were interpreted as lymph vessels or glandular ducts, as was also demonstrated by Bouma et al.18 whose overall observations were similar to ours. Kobayashi et al.10 found that the EP in vitro appeared as a relatively more reflective layer than the LP and MM, whereas in the present study exactly the opposite was demonstrated. In their review, Brand et al.13 stated that in vivo and in vitro OCT images differ considerably, and that the EP appeared more reflective than the LP in vitro. Nevertheless, the OCT images obtained in vitro with the SLD-based system in the present study closely resemble published in vivo images. Indeed, our observations closely agree with recent observations from 2 new groups who report that the EP is less reflective than the LP, and that the MM is less reflective than the LP and SM.14,20 Sivak et al.21 suggested that esophageal MM in vivo appears as a relatively thick triple layer structure with 2 highly reflective layers separated by one of lower reflectivity. A triple-layer structure in vitro with the layers adjacent to the MM appearing more reflective and the MM less reflective as mentioned earlier was also observed. Moreover, Sivak et al.21 stated that the SM in vivo appeared as relatively less reflective beneath this triple layer. In contrast, with the SLD-based OCT system used in the present study, the SM in vitro appeared as a relatively more reflective layer below the MM (Fig. 3C). This observation is in close agreement with other published data.12,20 Nonetheless, it is recognized that the saline wash and formalin fixation may have changed inherent optical properties and, consequently, the OCT appearance of the esophageal wall in ways that have not yet been determined. Thus, there is a need to corroborate the findings of the present in vitro study with clinical studies correlating in vivo OCT images with biopsy results. An OCT image is mathematically a convolution of light focus and coherence gating within tissue. In most OCT systems described to date, the light focus on a sample arm is stationary whereas the reference VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
focus is moved to achieve depth scanning. This may lead to conflicting data concerning the reflectivity of a particular tissue layer in an OCT image. In other words, depending on the location of the focus of the sample beam within the sample, the intensity of light detected at a particular depth may change. Consequently, the appearance of specific layers may differ, even though most investigators have used 1300 nm OCT systems for GI imaging. Therefore, it is our belief that by using dynamic focusing, that is, by moving sample and reference foci simultaneously, a more coherent and consistent appearance of tissue layers in OCT images will be obtained. ACKNOWLEDGEMENT The participation of Dr. Inci Çilesiz in this work was made possible by an official leave of absence from the Istanbul Technical University, Istanbul, Turkey. REFERENCES 1. Pera M, Cameron AJ, Trastek VF, Carpenter HA, Zinsmeister AR. Increasing incidence of adenocarcinoma of the esophagus and esophagogastric junction. Gastroenterology 1993;104:510-3. 2. Devesa SS, Blot WJ, Fraumeni JF. Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998;83:2049-53. 3. van Sandick JW, van Landschot JJB, Kuiken BW, Tytgat GNJ, Offerhaus GJ, Obertop H. Impact of endoscopic biopsy surveillance of Barrett’s esophagus on pathological stage and clinical outcome of Barrett’s carcinoma. Gut 1998;43:216-22. 4. Moretó M. Diagnosis of esophagogastric tumors. Endoscopy 2001;33:1-7. 5. Pfau RP, Sivak MV Jr. Endoscopic diagnostics. Gastroenterology 2001;120:763-81. 6. Geboes K, van Eyken P. The diagnosis of dysplasia and malignancy in Barrett’s oesophagus. Histopathology 2000;37:99-107. 7. Vieth M, Stolte M. Barrett’s mucosa, Barrett’s dysplasia and Barrett’s carcinoma: diagnostic endoscopy without biopsytaking does not suffice. Dis Esophagus 2000;13:23-7. 8. Das A, Sivak MV Jr, Chak A, Wong RCK, Westphal V, Rollins AM, et al. High-resolution endoscopic imaging of the GI tract: a comparative study of optical coherence tomography versus high-frequency catheter probe EUS. Gastrointest Endosc 2001;54:219-24.
VOLUME 56, NO. 6, 2002
I Çilesiz, P Fockens, R Kerindongo, et al.
9. Mallery S, van Dam J. Endoscopic practice at the start of the new millenium. Gastroenterology 2000;118:S129-47. 10. Kobayashi K, Izatt JA, Kulkarni MD, Willis J, Sivak MV Jr. High-resolution cross-sectional imaging of the gastrointestinal tract using optical coherence tomography: preliminary results. Gastrointest Endosc 1998;47:515-23. 11. Rollins AM, Ung-arunyawee R, Chak A, Wong RCK, Kobayashi K, Sivak MV Jr, et al. Real-time in vivo imaging of human gastrointestinal ultrastructure by use of endoscopic optical coherence tomography with a novel efficient interferometer design. Optics Letters 1999;24:1358-60. 12. Jäckle S, Gladkova N, Feldchtein F, Terentieva A, Brand B, Gelikonov G, et al. In vivo endoscopic optical coherence tomography of the human gastrointestinal tract—toward optical biopsy. Endoscopy 2000;32:743-9. 13. Brand S, Poneros JM, Bouma B, Tearney GJ, Compton CC, Nishioka NS. Optical coherence tomography in the gastrointestinal tract. Endoscopy 2000;32:796-803. 14. Li WD, Boppart SA, van Dam J, Mashimo H, Mutinga M, Drexler W, et al. Optical coherence tomography: advanced technology for the endoscopic imaging of Barrett’s esophagus. Endoscopy 2000;32:921-30. 15. Welch AJ, van Gemert MJC, Star W, Wilson BC. Definitions and overview of tissue optics. In: Welch AJ, van Gemert MJC, editors. Optical-thermal response of laser-irradiated tissue. New York: Plenum Press; 1995. p. 15-46. 16. Rollins AM, Kulkarni MD, Yazdanfar S, Ung-arunyawee R, Izatt JA. In vivo video rate optical coherence tomography. Optics Express 1998;3:219-29. 17. Stevens A, Lowe J. Alimentary tract. In: Human histology. 2nd ed. St. Louis: Mosby; 1997. p. 177-215. 18. Bouma BE, Tearney GJ, Compton CC, Nishioka NS. High-resolution imaging of the human esophagus and stomach in vivo using optical coherence tomography. Gastrointest Endosc 2000;51:467-74. 19. Tearney GJ, Brezinski ME, Southern JF, Bouma BE, Boppart SA, Fijumoto JG. Optical biopsy in human gastrointestinal tissue using optical coherence tomography. Am J Gastroenterol 1997;92:1800-4. 20. Zuccaro G, Gladkova N, Vargo J, Feldchtein F, Zagaynova E, Conwell D, et al. Optical coherence tomography of the esophagus and proximal stomach in health and disease. Am J Gastroenterol 2001;96:2633-9. 21. Sivak MV Jr, Kobayashi K, Izatt JA, Rollins AM, Ung-urunyawee R, Chak A, et al. High-resolution endoscopic imaging of the GI tract using optical coherence tomography. Gastrointest Endosc 2000;51:474-9.
GASTROINTESTINAL ENDOSCOPY
857
Background: Early diagnosis of esophageal cancer limited to the mucosa allows local endoscopic treatment and thereby improves prognosis. Optical coherence tomography images of normal human esophageal tissue obtained with 2 systems with light sources that provide different wavelengths (800 nm and 1275 nm) were compared with histology to determine which wavelength is best suited for detailed optical coherence tomography imaging of the esophageal wall, and to precisely localize the muscularis mucosae. Methods: Within 1 hour of surgical resection, an esophageal specimen was cleaned of excess blood with saline solution and soaked in formalin for a minimum of 48 hours. After optical coherence tomography imaging, the specimen was prepared for routine histologic assessment. To precisely localize the different layers of the esophageal wall on an optical coherence tomography image, well-defined structures within the esophageal wall were sought. Results: The 1275 nm system with 12 µm resolution was superior in terms of imaging depth. As compared with histology, the 4 µm resolution of the 800 nm system made fine detail more visible. With minimal experience, the muscularis mucosae could be recognized with either system as a hyporeflective layer with a diameter of around 180 µm. Conclusions: Based on appearance and location of morphologic landmarks, layers of normal esophageal wall, specifically, the location and extent of the muscularis mucosae, could be recognized by using both the 800 nm and 1275 nm optical coherence tomography system. Although different conditions may be operative in vivo, the present ex vivo study further verifies by precise interpretation that optical coherence tomography provides precise images of the esophageal wall. (Gastrointest Endosc 2002;56:852-7.)
The incidence of adenocarcinoma associated with Barrett’s esophagus is rising rapidly in the developed world.1,2 Accurate diagnosis plus intervention before cancer has invaded beyond the muscularis mucosae will hopefully improve prognosis for patients with this disease.3,4 Because existing methods are adequate for removal of early stage lesions in a minimally invasive manner, accurate in vivo delineation of dysplastic lesions and early stage canReceived December 28, 2001. For revision May 8, 2002. Accepted July 29, 2002. Current affiliations: Biomedical Engineering Program, Electronics and Communication Engineering Department, Istanbul Technical University, Istanbul, Turkey, and Department of Gastroenterology, Lasercentrum, and Department of Pathology, Academisch Medisch Centrum, Amsterdam, The Netherlands. Part of the research program of FOM [Stichting voor Fundamenteel Onderzoek der Materie, financially supported by NWO, the Nederlandse Organisatie voor Wetenschappelijk Onderzoek](IÇ, DF), and Philips Research (IÇ). This work was also funded by ZON [Zorg Onderzoek Nederland](RK). Reprint requests: Ton van Leeuwen, PhD, Lasercentrum, Meibergdreef 9, 1105AZ Amsterdam, The Netherlands. Copyright © 2002 by the American Society for Gastrointestinal Endoscopy 0016-5107/2002/$35.00 + 0 37/1/129606 doi:10.1067/mge.2002.129606 852
GASTROINTESTINAL ENDOSCOPY
cers is essential for selection of optimal treatment and minimizing morbidity. The current diagnostic method of choice, endoscopy with white light illumination, offers remarkable capabilities including noninvasive visual access to inner organs and direct tissue sampling. Nevertheless, visualization of only the luminal aspect of the mucosa under relatively low magnification places microscopic clues to underlying pathophysiology beyond the reach of the endoscopist.5 In general, few endoscopic features permit differentiation of nondysplastic and dysplastic mucosa. Moreover, early stage invasive cancer may not be visible endoscopically and may not be detected without biopsies, a method hampered by a substantial sampling error.3,6 A recent study involving histopathological diagnoses based on biopsy specimens from 1068 patients who underwent endoscopy because of gastroesophageal reflux disease found that 92% of cases of dysplasia and 33% of carcinomas in patients with Barrett’s esophagus were undetected at endoscopy by experienced endoscopists.7 EUS, even at high frequency (30 MHz), provides cross-sectional images with relatively low spatial resolution (100 µm at best).5,8,9 In contrast, optical VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
I Çilesiz, P Fockens, R Kerindongo, et al.
Figure 1. Simplified diagram of the fiberoptic Michelson interferometer used for OCT imaging. RSOD, Rapid scanning optical delay line.
coherence tomography (OCT), a newly emerging high resolution imaging technique for analysis of subsurface tissue structures to a maximum depth of 2 mm, is an excellent candidate method for determining noninvasively and in vivo the location, extent, and progression of normal and altered structures, including dysplasia, by guiding and supplementing endoscopy. Preliminary investigations have demonstrated the feasibility and superiority of OCT imaging systems with 1300 nm light sources for visualization of tissue layers.5,10-14 Analogous to B-mode US imaging, OCT imaging provides 2-dimensional cross-sectional morphologic images with spatial resolution approaching cellular level (i.e., ≤10 µm laterally and axially), leading to real-time visualization of tissue substructure that is nearly equivalent to standard microscopy of histologic sections. The characteristics of photon propagation in tissue, and consequently the attributes of OCT images, are primarily governed by the local optical properties of tissue: (1) absorption, (2) scattering coefficients, and (3) degree of scattering anisotropy of different layers of tissue at varying wavelengths of the optical spectrum.15 Theoretically, any morphologic formation unlike the normal tissue substructure will cause a local change in optical properties, particularly scattering coefficient and anisotropy, and will thus modify the optical behavior of tissue, which will in turn be detectable by OCT. Hence, OCT imaging has revolutionary potential to greatly minimize the sampling error associated with standard endoscopy and random biopsies and to open the possibility of early endoscopic diagnosis of dysplasia in situ. VOLUME 56, NO. 6, 2002
This study investigated the feasibility of using 2 laboratory OCT systems, based on different light sources (Ti:Sapphire laser, around 800 nm; a super luminescent diode, around 1275 nm), to characterize esophageal tissue layers with special emphasis on localization of the muscularis mucosae (MM). The study rationale was that cancer invasion beyond the MM may preclude local endoscopic treatment and thus confers a worse prognosis for patients. MATERIALS AND METHODS Tissue specimen An area of normal-appearing proximal esophagus was removed from the resection specimen from a patient who underwent operation for gastric cardia carcinoma. The tissue specimen was obtained within 1 hour after resection. Excess blood was removed by washing with saline solution and the muscularis propria was stripped off. The specimen was then stretched and pinned on clear agar gel in a Petri dish. The prepared specimen was soaked in formalin for a minimum of 48 hours for optimal preservation of tissue morphology. Before OCT imaging the specimen was submerged in normal saline solution. Imaging sites were marked on a drawing of the tissue sample (tissue map) for OCT and histology sections. The estimated correspondence among imaging sites and histology sections was ± 1 mm. After OCT imaging with both systems, imaging sites were prepared and stained with hematoxylin & eosin. Finally, the specific layers in histologic sections viewed under a standard light microscope were compared and contrasted with OCT images. OCT imaging Comparable OCT images of human esophageal tissue were acquired with one OCT system based on a commercially available femtosecond Ti:Sapphire laser (TSL) at GASTROINTESTINAL ENDOSCOPY
853
I Çilesiz, P Fockens, R Kerindongo, et al.
OCT imaging of human esophagus, localization of muscularis mucosae
Figure 2. OCT images and comparative histologic section of normal esophageal wall with an embedded ellipsoid structure. A, OCT image with TSL-based system. B, Photomicrograph of comparative histologic section showing lymphoid follicle (lf) embedded between lp and mm (H&E, orig. mag. ×40). C, OCT image with SLD-based system. ep, Epithelium; lp, lamina propria; mm, muscularis mucosae; sm, submucosa. 800 nm and a second OCT system based on a super luminescent diode (SLD) at 1275 nm. Technical specifications for the 2 systems were as follows: TSL system, optical bandwidth, 90 nm full width at half maximum (FWHM); axial resolution, less than 4 µm in tissue; optical power incident on sample/spot size, 4 mW/5 µm diameter; dynamic range (detection sensitivity), –115 dB. SLD system, optical bandwidth, 50 nm FWHM; axial resolution, approximately 12 µm in tissue; optical power incident on sample/spot size, 0.5 mW/20 µm diameter; dynamic range (detection sensitivity): –105 dB. Figure 1 is a simplified diagram of the optical fiber-based OCT systems. The probing/scanning depth was gated by low coherence interferometry with the tissue sample placed in the sample arm of a Michelson interferometer and a double pass rapid scanning optical delay (DP-RSOD) line in the reference arm.16 The DP-RSOD line was driven by a round-off triangular waveform at 50 Hz, allowing a scan depth of 1.5 mm. Lateral scanning was accomplished by translation of either the sample (TSL-based system) or the sample beam (SLD-based system). Analogous to US imaging, a two-dimensional OCT image (brightness or Bscan) was composed of a series of adjacent axial depth scans (amplitude or A-scans). In experiments with the TSL-based system, each 4 mm long B-scan consisted of 400 A-scans separated by 10 µm. Each 3.4 mm long B-scan in the SLD-based system consisted of 500 A-scans separated by 6.8 µm. Detected light intensity was optimized for image contrast with 256 gray levels. To facilitate the distinction between different layers of the esophageal wall (epithelium [EP], lamina propria [LP], MM, submucosa [SM]), and to precisely localize these layers on an OCT image, well-defined and easily identifiable features were sought that appeared repeatedly in a certain location within the esophageal wall. Imaging sites that had such landmark features were reimaged with 5 to 10 B-scans separated by 100 to 500 µm. Such sites were marked on the tissue map to guide histologic sectioning. Guided histologic sections were then studied to facilitate resolution of the location of the land854
GASTROINTESTINAL ENDOSCOPY
mark features within the esophageal wall, and to more accurately interpret OCT appearance of the layers adjacent to such landmark features.
RESULTS OCT was first performed on a total of 27 samples taken from the esophageal specimen with the TSLbased system. The layered structure of the esophageal wall was clearly identified in all images, the layers being distinguished by the relative differences in the intensity of their backscattered reflection. Based on a comprehensive review of all of these images, several well-defined dark ellipsoid structures embedded within the esophageal wall were discovered. After a second review of all OCT images and the acquisition of repeated OCT images from 8 of these imaging sites, a total of 4 imaging sites believed to contain the ellipsoid structures were selected for imaging with the SLD-based system. These ellipsoid structures were not visible to the naked eye with the specimen soaked in formalin or saline solution. However, while the samples were dried and prepared for histologic processing, clusters of the ellipsoid structures were identified over an area of approximately 2 cm2 just underneath the esophageal epithelium by using a low-power magnifying glass. Histologic sections of the well-defined ellipsoid structures revealed the typical appearance of lymphoid follicles. Examples of histologic and comparative OCT images obtained from such a site are shown in Figure 2. The follicles appeared darker (less reflective) than the surrounding structures in all OCT images and were 300 to 600 µm wide, 250 to 300 µm thick, and up to 500 µm long. Their repeated appearance between LP and MM facilitated precise localization of LP and MM and identification of VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
I Çilesiz, P Fockens, R Kerindongo, et al.
Figure 3. OCT images and comparative histologic section of normal esophageal wall: A, OCT image with TSL-based system. B, Photomicrograph of comparative histologic section (H&E, orig. mag. ×40). C, OCT image with SLD-based system. ep, Epithelium; lp, lamina propria; mm, muscularis mucosa; sm, submucosa.
the first 3 layers of the esophageal wall (EP, LP, MM) in all imaged sections. In the positive (commonly referred to as grayscale) OCT image display modality used for both systems, bright white corresponds to highly reflective whereas dark gray or black corresponds to less reflective. An additional histologic section and OCT images obtained with both OCT systems are shown in Figure 3. In OCT images of normal esophageal wall, desquamating cells appeared as a thin bright (highly reflective) line. The EP appeared darker (less reflective), and the LP slightly brighter (more reflective). The interface between EP and LP was relatively easier to recognize on OCT images obtained with the SLDbased system. MM appeared darker (less reflective) than LP and also “wavy” on images acquired with either system. However, the wavy pattern was easily recognized on images obtained with the TSL-based system, but not as obvious in those with the SLDbased system (Fig. 3A and C). Nevertheless, the boundaries of the MM could be identified by “lines” running horizontally above and under the MM in images acquired with both systems. These lines indicated change of structure and scattering properties, that is, interfaces with LP and SM. From a total of 77 B-scans with and without follicles present, the MM was identified as a 180 ± 22 µm thick hyporeflective layer. Boundaries of the MM could additionally be identified by the brighter (more reflective) appearance of the SM in OCT images obtained with the SLD-based system; whereas for images acquired with the TSL-based system, “visibility” was greatly reduced beyond the MM and the SM appeared dark. Loss of visibility at depths greater than 500 to 600 µm in the tissue (Fig. 3A) may be caused by relatively shallower penetration of light and higher scattering properties at 800 nm compared with 1275 nm. In VOLUME 56, NO. 6, 2002
general, more detail could be observed in images acquired with the TSL-based system, as anticipated based on the 3-fold greater resolution of that system. DISCUSSION To our knowledge, this is the first comparison study of histology versus OCT images from the normal esophageal wall with 2 different systems with 2 different wavelengths, in which distinct, well-defined morphologic markers such as lymphoid follicles are shown. The LP throughout the digestive tract contains cells belonging to the immune system. Lymphoid cells are commonly arranged as large follicles partly within the mucosa and partly within the SM, splitting the MM.17 In contrast, the esophageal LP contains scattered lymphocytes. Nevertheless, aggregations of lymphoid cells forming small follicles are common in the esophageal SM near the squamous/columnar epithelial junction and occasionally in upper parts of normal esophageal SM.17 Consequently, the frequent presence of lymphoid follicles in the upper histologic layers of the esophageal wall away from the squamous/columnar epithelial junction was considered a manifestation of the inflammation/immune response to chronic gastroesophageal reflux. Although only one esophagus specimen was included in this study, OCT images were examined from 3 different pathologic specimens from 3 different patients and similar follicles were observed. The time required for preparation was too long for 2 of these specimens, so that tissue decomposition could influence the optical properties of the different tissue layers; thus, these 2 specimens were not included in the study. Note that in histologic sections, periesophageal mucous glands may appear similar to follicles (Fig. 2). The different layers of normal esophageal wall could be precisely identified with both the 800 nm– GASTROINTESTINAL ENDOSCOPY
855
I Çilesiz, P Fockens, R Kerindongo, et al.
and 1275 nm–based systems by comparison with corresponding histologic sections. These results indicate that the SLD-based OCT system with 12 µm axial resolution was superior in terms of penetration and visibility of deeper layers compared with the TSL-based system. Based on analysis of the histologic findings, the 4 µm resolution of the TSLbased system made fine detail and structures clearly recognizable. Although a better resolution with the SLD-based system would improve visualization of detail, with minimal experience the interface between LP and MM could be recognized by using either system. In addition to increased imaging depth, an SLD-based system with slightly less resolution may be preferable for clinical esophageal OCT imaging because it is portable, less costly, and easier to maintain when compared with a bulkier, more expensive, TSL-based system. Other groups have demonstrated the feasibility of endoscopic high-resolution OCT imaging in the GI tract.5,12-14 However, these investigators have compared histologic sections with OCT images acquired with 1300 nm systems of varying optical bandwidths. Although separation between SM and muscularis propria as a result of saline solution injection has been captured on an OCT image,13 no morphologically well defined structures were sought as a means to interpret OCT images and localize the MM. In general, OCT images and histologic sections from cadaveric or autopsy specimens have been used to study the layered structure of the human esophageal wall and to interpret OCT images obtained in vivo.18 In the first report of GI OCT imaging, autopsy specimens were used to study esophageal structures.19 By using negative (i.e., black on white) OCT imaging, these investigators stated that MM appeared brighter than the mucosa and that gaps were observable between the MM and SM. In a positive OCT image display, also referred to as grayscale (white-on-black), “brighter” or “whiter” signifies “more reflective,” whereas “darker” or “grayer” signifies “less reflective.” In contrast, in a negative OCT display, also referred to as inverse grayscale (black-on-white), “brighter” signifies “less reflective,” whereas “darker” signifies “more reflective.” To avoid confusion, an independent image display terminology was chosen, and the terms “more reflective” and “less reflective” were used. The boundaries of the MM as observed in the present study appeared more reflective. However, the MM was in general less reflective than the LP and appeared “wavy,” an observation not previously reported. The explanation for this observation is believed to be two-fold: the relatively higher resolution (4 µm vs. 10-20 µm in earlier studies) made fine 856
GASTROINTESTINAL ENDOSCOPY
OCT imaging of human esophagus, localization of muscularis mucosae
detail more visible, and by stretching and pinning the sample on clear agar gel, the EP appeared to be 100 to 150 µm thick, which is much thinner than the 500 to 800 µm previously reported.17 Stretching the sample also flattened the irregular lower border of the EP. Therefore, back-scattered photons of light coming from deeper layers of the esophageal wall might have had a better chance to propagate toward the lumen without being scattered by irregular borders. The flattening of EP may have thus “improved” detection sensitivity for signals coming from deeper layers, thereby providing greater structural detail. Gaps were frequently observed between and within the layers of the esophageal wall. These were interpreted as lymph vessels or glandular ducts, as was also demonstrated by Bouma et al.18 whose overall observations were similar to ours. Kobayashi et al.10 found that the EP in vitro appeared as a relatively more reflective layer than the LP and MM, whereas in the present study exactly the opposite was demonstrated. In their review, Brand et al.13 stated that in vivo and in vitro OCT images differ considerably, and that the EP appeared more reflective than the LP in vitro. Nevertheless, the OCT images obtained in vitro with the SLD-based system in the present study closely resemble published in vivo images. Indeed, our observations closely agree with recent observations from 2 new groups who report that the EP is less reflective than the LP, and that the MM is less reflective than the LP and SM.14,20 Sivak et al.21 suggested that esophageal MM in vivo appears as a relatively thick triple layer structure with 2 highly reflective layers separated by one of lower reflectivity. A triple-layer structure in vitro with the layers adjacent to the MM appearing more reflective and the MM less reflective as mentioned earlier was also observed. Moreover, Sivak et al.21 stated that the SM in vivo appeared as relatively less reflective beneath this triple layer. In contrast, with the SLD-based OCT system used in the present study, the SM in vitro appeared as a relatively more reflective layer below the MM (Fig. 3C). This observation is in close agreement with other published data.12,20 Nonetheless, it is recognized that the saline wash and formalin fixation may have changed inherent optical properties and, consequently, the OCT appearance of the esophageal wall in ways that have not yet been determined. Thus, there is a need to corroborate the findings of the present in vitro study with clinical studies correlating in vivo OCT images with biopsy results. An OCT image is mathematically a convolution of light focus and coherence gating within tissue. In most OCT systems described to date, the light focus on a sample arm is stationary whereas the reference VOLUME 56, NO. 6, 2002
OCT imaging of human esophagus, localization of muscularis mucosae
focus is moved to achieve depth scanning. This may lead to conflicting data concerning the reflectivity of a particular tissue layer in an OCT image. In other words, depending on the location of the focus of the sample beam within the sample, the intensity of light detected at a particular depth may change. Consequently, the appearance of specific layers may differ, even though most investigators have used 1300 nm OCT systems for GI imaging. Therefore, it is our belief that by using dynamic focusing, that is, by moving sample and reference foci simultaneously, a more coherent and consistent appearance of tissue layers in OCT images will be obtained. ACKNOWLEDGEMENT The participation of Dr. Inci Çilesiz in this work was made possible by an official leave of absence from the Istanbul Technical University, Istanbul, Turkey. REFERENCES 1. Pera M, Cameron AJ, Trastek VF, Carpenter HA, Zinsmeister AR. Increasing incidence of adenocarcinoma of the esophagus and esophagogastric junction. Gastroenterology 1993;104:510-3. 2. Devesa SS, Blot WJ, Fraumeni JF. Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998;83:2049-53. 3. van Sandick JW, van Landschot JJB, Kuiken BW, Tytgat GNJ, Offerhaus GJ, Obertop H. Impact of endoscopic biopsy surveillance of Barrett’s esophagus on pathological stage and clinical outcome of Barrett’s carcinoma. Gut 1998;43:216-22. 4. Moretó M. Diagnosis of esophagogastric tumors. Endoscopy 2001;33:1-7. 5. Pfau RP, Sivak MV Jr. Endoscopic diagnostics. Gastroenterology 2001;120:763-81. 6. Geboes K, van Eyken P. The diagnosis of dysplasia and malignancy in Barrett’s oesophagus. Histopathology 2000;37:99-107. 7. Vieth M, Stolte M. Barrett’s mucosa, Barrett’s dysplasia and Barrett’s carcinoma: diagnostic endoscopy without biopsytaking does not suffice. Dis Esophagus 2000;13:23-7. 8. Das A, Sivak MV Jr, Chak A, Wong RCK, Westphal V, Rollins AM, et al. High-resolution endoscopic imaging of the GI tract: a comparative study of optical coherence tomography versus high-frequency catheter probe EUS. Gastrointest Endosc 2001;54:219-24.
VOLUME 56, NO. 6, 2002
I Çilesiz, P Fockens, R Kerindongo, et al.
9. Mallery S, van Dam J. Endoscopic practice at the start of the new millenium. Gastroenterology 2000;118:S129-47. 10. Kobayashi K, Izatt JA, Kulkarni MD, Willis J, Sivak MV Jr. High-resolution cross-sectional imaging of the gastrointestinal tract using optical coherence tomography: preliminary results. Gastrointest Endosc 1998;47:515-23. 11. Rollins AM, Ung-arunyawee R, Chak A, Wong RCK, Kobayashi K, Sivak MV Jr, et al. Real-time in vivo imaging of human gastrointestinal ultrastructure by use of endoscopic optical coherence tomography with a novel efficient interferometer design. Optics Letters 1999;24:1358-60. 12. Jäckle S, Gladkova N, Feldchtein F, Terentieva A, Brand B, Gelikonov G, et al. In vivo endoscopic optical coherence tomography of the human gastrointestinal tract—toward optical biopsy. Endoscopy 2000;32:743-9. 13. Brand S, Poneros JM, Bouma B, Tearney GJ, Compton CC, Nishioka NS. Optical coherence tomography in the gastrointestinal tract. Endoscopy 2000;32:796-803. 14. Li WD, Boppart SA, van Dam J, Mashimo H, Mutinga M, Drexler W, et al. Optical coherence tomography: advanced technology for the endoscopic imaging of Barrett’s esophagus. Endoscopy 2000;32:921-30. 15. Welch AJ, van Gemert MJC, Star W, Wilson BC. Definitions and overview of tissue optics. In: Welch AJ, van Gemert MJC, editors. Optical-thermal response of laser-irradiated tissue. New York: Plenum Press; 1995. p. 15-46. 16. Rollins AM, Kulkarni MD, Yazdanfar S, Ung-arunyawee R, Izatt JA. In vivo video rate optical coherence tomography. Optics Express 1998;3:219-29. 17. Stevens A, Lowe J. Alimentary tract. In: Human histology. 2nd ed. St. Louis: Mosby; 1997. p. 177-215. 18. Bouma BE, Tearney GJ, Compton CC, Nishioka NS. High-resolution imaging of the human esophagus and stomach in vivo using optical coherence tomography. Gastrointest Endosc 2000;51:467-74. 19. Tearney GJ, Brezinski ME, Southern JF, Bouma BE, Boppart SA, Fijumoto JG. Optical biopsy in human gastrointestinal tissue using optical coherence tomography. Am J Gastroenterol 1997;92:1800-4. 20. Zuccaro G, Gladkova N, Vargo J, Feldchtein F, Zagaynova E, Conwell D, et al. Optical coherence tomography of the esophagus and proximal stomach in health and disease. Am J Gastroenterol 2001;96:2633-9. 21. Sivak MV Jr, Kobayashi K, Izatt JA, Rollins AM, Ung-urunyawee R, Chak A, et al. High-resolution endoscopic imaging of the GI tract using optical coherence tomography. Gastrointest Endosc 2000;51:474-9.
GASTROINTESTINAL ENDOSCOPY
857