Journal of the Korean Physical Society, Vol. 57, No. 1, July 2010, pp. 79∼83

Feasibility of Full-field Optical Coherence Microscopy in Ultra-structural Imaging of Human Colon Tissues Eun Seo Choi Department of Physics, Chosun University, Gwangju 501-759

Woo June Choi, Seon Young Ryu and Byeong Ha Lee Department of Information and Communications, Gwangju Institute of Science and Technology, Gwangju 500-712

Jae-Hyuk Lee Department of Pathology, Chonnam National University Hospital, Gwangju 501-757

Hee-Seung Bom and Byeong Il Lee∗ Department of Nuclear Medicine, Chonnam National University Hospital, Gwangju 501-757 (Received 21 June 2010) We demonstrated the imaging feasibility of full-field optical coherence microscopy (FF-OCM) in pathological diagnosis of human colon tissues. FF-OCM images with high transverse resolution were obtained at different depths of the samples without any dye staining or physical slicing, and detailed microstructures of human colon tissues were visualized. Morphological differences in normal tissues, cancer tissues, and tissues under transition were observed and matched with results seen in conventional optical microscope images. The optical biopsy based on FF-OCM could overcome the limitations on the number of physical cuttings of tissues and could perform high-throughput mass diagnosis of diseased tissues. The proved utility of FF-OCM as a comprehensive and efficient imaging modality of human tissues showed it to be a good alternative to conventional biopsy. PACS numbers: 42.62.Be, 87.63.+n, 42.87.Bg Keywords: FF-OCM, Human colon tissues, Optical biopsy DOI: 10.3938/jkps.57.79

performance. A series of interference fringes obtained from LCI is recorded at a photodetector and analyzed at sites of the interfaces along depth of specimen, where the intensity of the reflected signal is proportional to the refractive index difference around the interfaces. If a mechanical scanner for beam steering is added to the sample arm, OCT can obtain a multi-dimensional highresolution tomographic view. Conventional interferometers generate interferograms with modulating OPD by using an optical delay line, and the interferogram is registered as a function of the OPD. This method is typically called the time-domain detection scheme. When this scheme is applied to OCT (timedomain OCT), imaging performance, such as imaging speed and sensitivity, strongly depends on the operational conditions of the optical delay line. Due to mechanical operation, real-time imaging with large probing depth is only available under restricted conditions. The problem can be properly solved by adopting a frequencydomain detection scheme [4,5], which measures a spec-

I. INTRODUCTION Optical coherence tomography (OCT) provides highresolution anatomic images of a specimen without the physical slicing required in conventional biopsy [1]. The working principle of OCT for optical biopsy is based on low-coherence interferometry (LCI) [2,3], which employs a broadband light source in an interferometer to utilize its low coherence property. Since interference fringes are available when the optical path-length difference (OPD) between the sample arm and the reference arm is less than the coherence length given by the sources, the low coherence characteristics in LCI greatly reduces the full width at half maximum (FWHM) of the interferogram to around few microns. The FWHM is used as criteria for the minimally-differentiable distance or feature size of the sample in an OCT image, and a decrease in the FWHM is directly linked to high-resolution imaging ∗ E-mail:

[email protected]

-79-

-80-

trogram or spectral interference fringes instead of an interferogram. Because the spectrogram is measured in the frequency-domain not the time-domain, OPD information on the sample is deduced by taking an inverse Fourier transform of the spectrogram. Measurement of the spectrogram is performed without moving the optical delay line, which can improve the imaging speed at a low noise level. OCT based on the frequency-domain detection scheme (FD-OCT [6,7]), which exploits a highspeed wavelength-swept source and a single detector [8] or a broadband source and a high-speed charge coupled device (CCD) camera for parallel detection [9], is widely used to secure improved image quality. FD-OCT achieves fast imaging of human organs in vivo [10], and dynamic phenomena in human body [11] are properly observed. Feasibility studies of the OCT technique in human tissue imaging were carried out and are reported in various literatures. Among them, the imaging of human colon tissues was extensively carried out [12–14]. The morphology of colon crypts on the surface of normal colorectal mucosa shows regular pits pattern having roundish openings in a well-defined direction. However, this characteristic shape in normal colon tissue is changed into an irregular shape when developing to a cancerous state. Previous studies demonstrated the high relevance between the morphological change of colon crypts and the degree of progress in colon disease [15,16]. Therefore, OCT imaging is readily applicable to in-situ diagnosis of colon tissues. By imaging the morphological changes of colon crypts, such as breaking of circular symmetry and increased randomness in orientation, the utility of OCT imaging for evaluating colonic disease progression was proven [17]. The morphological characteristics across colon crypts are maintained along the depth of the colon tissues. Optical diagnostic of colon tissues requires depthresolved high-transverse-resolution en-face imaging and high-axial-resolution imaging. Although OCT achieves high-axial-resolution depth-resolved optical imaging, the transverse resolution in the en-face imaging is poor. This low transverse resolution is not applicable to detailed descriptions of the colon tissues. The low axial resolution of OCT can be further enhanced by using optical sources having wider bandwidths, but the transverse resolution is not further improved because of the diffraction limit of the objective lens used. A confocal imaging technique [18] may be a promising alternative to overcome the low transverse resolution in optical imaging. When confocal microscopy is applied, the imaging depth is less than a few hundred microns, which corresponds to the depth of a shallow epithermal layer. If useful information about the colon crypts is to be obtained, the probing depths should be extended to depth larger than 1 mm. The adaptation of a microscopic technique to OCT imaging is a good alternative to hurdle the problems of the low transverse resolution in OCT and the short depth range of the conical technique simultaneously. That imaging

Journal of the Korean Physical Society, Vol. 57, No. 1, July 2010

method is called full-field optical coherence microscopy (FF-OCM) [19]. FF-OCM performs optical imaging with both high transverse resolution of less than 1 micron and high axial resolution in OCT at the same time. The working principle of FF-OCM is similar to that of OCT. Therefore, FF-OCM imaging does not need a process of dye staining or physical slicing of a specimen, where these are indispensible for microscope observation. In the research, we demonstrated the feasibility of using the FF-OCM technique in imaging human colon crypts at high transverse resolution. The detailed morphological characteristics in the axial and the transverse tomographic images were imaged with proving good correlation to histology. The transition from a normal to a cancerous state was also imaged clearly. The obtained FF-OCM images were compared with the histologies of the same specimens and presented good agreement.

II. EXPERIMENTS AND DISCUSSION Human colon tissues were resected from human colons incised during surgery. Freshly extracted colon specimens were washed using 10% acetylcysteine, a mucolytic agent, and were then washed with water to remove the coating mucous and to prepare the tissue for dye staining. The prepared specimen was fixed in a paraffin solution for physical slicing. FF-OCM imagings of regions of interest (ROI) marked for microscopic observation were performed before the fixing and slicing procedure. FFOCM observation was carried out by changing imaging depth. After FF-OCM imaging, the specimens were stained with 10% methylene blue to enhance the imaging contrast of crypts under the microscope. After dye staining, the specimens were sliced, and microscopic images at the marked ROI’s were taken under different magnifications by using a commercial CCD camera (PixelLink model PLA642). An evaluation of the correlation between the optical images acquired from different imaging modalities was carried out. A layout of our FF-OCM set-up is shown in Fig. 1. The system is based on Linnik type Michelson interferometry, which is explained in a previous publication [20]. An incoherent light beam from a white light illuminator of a 100-W halogen bulb is incident on a conventional bulk optic Michelson interferometer through a multimode imaging guider. The light beam launched into the fiber bundle illuminates a beam splitter homogenously with the aid of an adaptive optic system. In both interferometer arms, identical water immersion objectives (40×, 0.8-NA, Olympus) are employed for focusing and collecting backscattered light beams from scattering media and a reference mirror surface. A neutral density filter (NDF) is inserted into the reference arm to enhance the visibility of the interferogram. To compensate for the material dispersion imbalance due to NDF, we place window glass plate of the same material was placed in

Feasibility of Full-field Optical Coherence Microscopy in Ultra-structural Imaging · · · – Eun Seo Choi et al.

-81-

interferogram on CCD camera can be written as I(ax , ay , t) = Iincoherent (ax , ay ) +Icoherent cos[ϕ(ax , ay ) +ψ sin(ωt + θ)],

(2)

where ψ, θ and ω are the amplitude, initial phase, and angular frequency of the modulation, respectively. PZT in the reference mirror induces the phase modulation. The amplitude ψ is adjusted with the voltage applied to the PZT, and the initial phase θ is determined by choosing the trigger time of CCD image acquisition with respect to the PZT position. The interference signal is accumulated in the CCD over one-quarter of the overall modulation period. Four images obtained every one-quarter period are reconstructed into one en-face image at a fixed depth of the specimen. One image obtained in one period can be expressed as Fig. 1. Schematic of FF-OCM (BF: bandpass filter; NDF: neutral density filter; BS: beam splitter; WG: window glass; MO: microscope objective; RM: reference mirror; AM: angled mirror; PZT: piezo-actuator transducer; L: lens; CCD: charge coupled device)

In =

T Iincoherent 4 Z +Icoherent

nT /4

dt cos[φ + ψ sin(ωt + θ)],

(n−1)R/4

the sample arm. Retro-reflected beams are recombined by using the beam splitter forming the interferogram. The interferogram is detected and digitized by using a 12-bit CCD camera (CCD-1020, 512 × 512 pixels, 20 fps, VDS Vosskuhler). The registered intensity signal on the CCD camera is manipulated at a personal computer and are displayed using the Labview program (National Instrument, Ver. 6.5). A phase-shifting interferometric technique was applied to the interferometer to reconstruct an en-face image from the interferogram [21]. By electro-mechanical stepping of PZT, we controlled the phase modulation of the backscattered light in the reference arm with repeating periodic round-trip motion. The theoretical principle of image reconstruction in FF-OCM with integrating buckets is as follows [22]: When the input beam is monochromatic and is taken to be a plane wave, the interference pattern detected on the CCD camera is given by [23]

I(ax , ay ) = Iincoherent (ax , ay ) + Icoherent cos ϕ(ax , ay ), (1) where Iincoherent (ax , ay ) is the non-interfered optical signal and and ϕ (ax , ay ) is the optical path-length difference between the sample arm and the reference arm of the interferometer. The intensity of the coherent signal or the interference signal is denoted as Icoherent (ax , ay ) in the above equation. When the optical path-length of the reference arm is modulated sinusodially with the PZT, the phase of the

n = 0, 1, 2, 3, 4. (3) After many different kinds of mathematical manipulations of the above equation, only the coherent signal is extracted from the successively acquired CCD images. The optimum parameters used in the experiment are obtained by repeating the procedure in the reference [23]. The modulation amplitude and the phase were 3.47 and 1.02, respectively. Figure 2(a) shows H&E stained normal colon tissues cut across the crypts. A microscope view of the marked ROI is presented in Fig. 2(b), where the circular shape on a plane section of crypts is a typical feature of normal colon tissue. From Fig. 2(c), the circular shape on the plane section across the colon crypts was shown to be maintained at different depths in the colon tissues. The images also showed that the crypts were surrounded by lamina propria. Goblet cells at the center of the crypts were observed in FF-OCM images shown in Fig. 2(c), as in Fig. 2(b). Figure 2(b) was obtained after staining and slicing as in conventional biopsy, but the FF-OCM images listed in Fig. 2(c) were obtained at different depths before treatment of the sample. The FF-OCM images were obtained every 2 µm along the depth from -12 um to -20 um; the imaging depths are designated in the right lower corner of each figure. Cross-sectional views of human colon tissues cut across an inclined plane are shown in Figs. 3(a) and 3(b). Different ROIs in the same specimen were designated by arrows and were imaged with both a microscope and the FF-OCM system. The detailed microscopic structures of the sample, obtained with the microscope, are shown in Fig. 3(b). A cross-sectional view of the normal colon

-82-

Journal of the Korean Physical Society, Vol. 57, No. 1, July 2010

Fig. 2. (Color online) Normal human colon tissues cut across the crypts: (a) H&E stained slide image, (b) microscopy image (200X), and (c) FF-OCM images at different depths.

Fig. 4. (Color online) Normal human colon tissues cut longitudinally: (a) & (c) microscopy images (200X), (b) FFOCM image at a depth of -18 µm , (d) FF-OCM images at a depth of -17 µm.

Fig. 3. (Color online) Normal human colon tissues cut across an inclined plane: (a) H&E stained slide image, (b) microscopy image (200X), and (c) FF-OCM images at different depths.

tissues at different depths appears. The goblet cells and the lamina propria around the crypts are clearly shown in the obtained FF-OCM images in Fig. 3(c). The geometrical characteristic of the crypts shown in the FF-OCM images was maintained along the depth of the sample. Figure 4 presented imaging results of normal colon tissues cut longitudinally. The morphology of the colon crypts is clearly presented in Figs. 4(a) and 4(c), where the surface epithelium, a crypt, a crypt lumen and goblet cells are clearly imaged. The lamina propria between the epithelia of adjacent crypts was also imaged. The FF-OCM system also imaged distinct characteristics of the morphology of normal colon tissues. The discrepancy between the microscopic images and the FF-OCM images was caused by rotation of the sample when the images were obtained with a microscope. Contrary to normal colon tissues, colon cancer tissues do not show the characteristic morphological property. The imaging site is shown by an arrow in Fig. 5(a). The boundary of the surface epithelium is not clearly, and no crypts or goblet cells are seen in the microscope and the FF-OCM images. The FF-OCM images obtained are well matched with the microscopic images. At deep depths below 21 µm from the surface of the sample,

Fig. 5. (Color online) Human colon cancer tissues: (a) H&E stained slide image, (b) microscopy image (200X), and (c) FF-OCM images at different depths.

the overall disorder presented in the microscope image was similarly observed in the FF-OCM images. Tissues seem to be merged into one, without showing any obvious boundary of crypts. Figure 6 was obtained at the boundary between normal and cancer tissues. The tissue experienced a transition from normal tissue to cancer tissue. At a glance, this figure is similar to Fig. 2 and Fig. 3. Compared with normal tissues, in the colon tissues of Fig. 6(b), the goblet cells are not regularly shaped, and the size is increased. The characteristic circular shape of the colon crypts was transformed into lengthy ones. In the lower part of the figure, the boundary between the epithelia of the crypts and the laminar propria is not clear and is ambiguous. The well-defined structure of the crypts was changed to a poor one, losing all characteristics. In Fig. 6(c), the transition shown in Fig. 6(b) is easily observed. Compared with the images in Fig. 3(c), a morphologi-

Feasibility of Full-field Optical Coherence Microscopy in Ultra-structural Imaging · · · – Eun Seo Choi et al.

-83-

ACKNOWLEDGMENTS This work was supported in part by the Korea Science and Engineering Foundation (KOSEF) grant founded by the Korea government (MEST) (No. R01-2009-00020991-0) and by a grant from the Institute of Medical System Engineering (iMSE) in the Gwangju Institute of Science and Technology (GIST), Korea. Fig. 6. (Color online) Intermediate region between normal tissues and cancer colon tissues: (a) H&E stained slide image, (b) microscopy image (200X), and (c) FF-OCM images at different depths.

cal variation is certainly observed. Contrary to cancer tissues, the structure of the tissues was maintained, but detailed morphological characteristic could hardly be observed in the FF-OCM images in Fig. 3(c).

III. CONCLUSION In conclusion, we demonstrated the high potential of FF-OCM in imaging human colon tissues under different progression states from normal tissues to cancer tissues. The microstructures of the human colon tissues clearly presented two-dimensional en-face images when using FF-COM technique. Although a microscope view with a sliced sample can provide a detailed visualization of the sample, only limited sectioning of the sample is permitted. Compared with this method, FF-OCM allows repeated optical sectioning of the specimen without physical cutting or staining with offering similar imaging performance. Therefore, the proposed method overcomes the disadvantage of a limiting sectioning of the sample without degradation of the quality in the microscopic investigation. In addition, the high transverse-resolution of the FF-OCM presents detailed morphological information on human colon tissues such as epithelia, crypts, and goblet cells, as in microscopic view. With these merits, FF-OCM also was used to image transition tissues, as well as cancer tissues. The obtained FF-OCM images proved to have a close correlation to histology. This factor shows that the proposed method can visualize geometrical or morphological variations in human colon tissues and that property can be applied to diagnose the progression of diseases. FF-OCM could be a help in conducting rapid diagnosis of tissue samples extracted for the investigation of disease. When high-speed scanning techniques were employed to our proposed method, the en-face imaging scheme could enable mass investigation for medical diagnosis, replacing excisional biopsy.

REFERENCES [1] D. Huang et al., Science 254, 1178 (1991). [2] R. C. Youngquist, S. Carr and D. E. N. Davies, Opt. Lett. 12, 158 (1987). [3] K. Takada, I. Yokohama, K. Chida and J. Noda, App. Opt. 26, 1603 (1987). [4] F. Fercher, C. K. Hitzenberger, G. Kamp and S. Y. ElZaiat, Opt. Commun. 117, 43 (1995). [5] F. Lexer, C. K. Hitzenberger, A. F. Fercher and M. Kulhavy, Appl. Opt. 36, 6548 (1997). [6] G. Hausler and M. W. Lindner, J. Biomed. Opt. 3, 21 (1998). [7] S. R. Chinn, E. Swanson and J. G. Fujimoto, Opt. Lett. 22, 340 (1997). [8] S. Yun, G. Tearney, Johannes de Boer, N. Iftimia and B. Bouma, Opt. Express 11, 2953 (2003). [9] N. Nassif, B. Cense, B. H. Park, S. H. Yun, T. C. Chen, B. E. Bouma, G. J. Tearney and J. F. de Boer, Opt. Lett. 29, 480 (2004). [10] G. J. Tearney, M. E. Brezinski, B. E. Bouma, S. A. Boppart, C. Pitris, J. F. Southern and J. G. Fujimoto, Science 276, 2037 (1997). [11] Siavash Yazdanfar, Manish Kulkarni and Joseph Izatt, Opt. Express 1, 424 (1997). [12] P. R. Pfau et al., Gastrointest. Endosc. 58, 196 (2003). [13] D. C. Adler, Y. Chen, R. Huber, J. Schmitt, J. Connolly and J. G. Fujimoto, Nat. Photonics 1, 709 (2007). [14] S. H. Yun et al., Nat. Med. 12, 1429 (2006). [15] S. Kudo, C. A. Rubio, C. R. Teixeira, H. Kashida and E. Kogure, Endoscopy 33, 367 (2001). [16] S. Tamura, Y. Furuya, T. Tadokoro, Y. Higashidani, Y. Yokoyama, K. Arak and S. Onishi, J. Gastroenterol. 37, 798 (2002). [17] Xin Qi et al., J. Biomed. Opt. 13, 054055 (2008). [18] A. Polglase, W. McLaren, S. A. Skinner, R. Kiesslich, M. F. Neurath and P. Delaney, Gastrointest. Endosc. 62, 686 (2005). [19] A. Dubois, L. Vabre, A. C. Boccara and E. Beaurepaire, Appl. Opt. 41, 805 (2002). [20] Jihoon Na, Woo June Choi, Eun Seo Choi, Seon Young Ryu and Byeong Ha Lee, Appl. Opt. 47, 459 (2008). [21] A. Dubois, K. Grieve, G. Moneron, R. Lecaque, L. Vabre and C. Boccara, Appl. Opt. 43, 2874 (2004). [22] E. Beaurepaire, A. C. Boccara, M. Lebec, L. Blanchot and H. Saint-Jalmes, Opt. Lett. 23, 244 (1998). [23] A. Dubois, J. Opt. Soc. Am. A 18, 1972 (2001).