We present a polarization-sensitive optical coherence tomography (PS-OCT) technique that can quantify the polarization changes (the degrees of circular polarization, DOCP) caused by the scattering changes induced by cervical intraepithelial neoplasia (CIN). The axial and lateral resolutions of our PS-OCT system are 13 µm and 15 µm, respectively. Uterine cervical conization tissue samples from 18 patients were examined, and 71 areas were imaged for in vitro studies; about 2–4 areas per sample were imaged and processed for diagnosis. The scanned areas had a size of 2 mm (axial) × 2 mm (lateral) × 4 mm (transversal). We quantified the slope of the axial decay of the DOCP signal near the cervical epithelium by a linear fitting procedure. The excised samples were then investigated by two pathologists, and their histological findings were later compared with the PS-OCT results. Our results show that the sensitivity and specificity are 94.7% and 71.2%, respectively.
©2008 Optical Society of America
Worldwide, cervical cancer is the second most common malignancy in women, and it remains a leading cause of cancer-related death for the women in developing countries . It is important not only to detect cancer but also to discriminate the degree of cervical intraepithelial neoplasia (CIN-I, II, III), which is cervical dysplasia at the squamo-columnar junction. Low-grade squamous intraepithelial lesions (L-SIL) correspond to CIN-I, and highgrade SIL (H-SIL) to CIN-II, CIN-III, and carcinoma in situ (CIS). These stages can be differentiated only through a pathological examination. The screening of cervical pathology is usually performed by Papanicolaou (Pap) smear or colposcopy. Because Pap smears have low sensitivity (≤58%) and specificity (69%), many lesions, particularly those occurring at the sub-surface, are missed or overcalled ; this low sensitivity can lead to incorrect diagnosis. In addition, although expert colposcopy offers a high sensitivity of 96%, it still has a low specificity of 48% , which can lead to many unnecessary biopsies, resulting in higher costs. Recently, in vivo optical imaging technologies such as elastic-scattering spectroscopy , near-infrared (NIR) spectroscopy , fluorescence spectroscopy [5–7], confocal microscopy [8, 9], and optical coherence tomography [10, 11] have been studied for detecting early pathological changes in the cervix.
It is clear that high-resolution tomographic techniques can facilitate early-stage diagnosis of the cervix to a significant degree. Optical coherence tomography (OCT) is a relatively new noninvasive imaging technique that allows the acquisition of high-resolution tomographic images [10–12]. Previous studies have reported that OCT can detect the disruption of the basement membrane in the cervix caused by cancer invasion . However, we have found that the basement membrane may not be visible in the OCT image even when the cervix is not affected by any pathological condition . In another study, it has been reported that the averaged OCT signal slope near the epithelial region could be used for distinguishing between normal and abnormal cervical tissues . However, it is also known that the averaged intensity slope of OCT images can be significantly affected by the location of the focal point .
It is well known that when CIN is present in the cervix, the nuclear/cytoplasmic (N/C) area ratio increases significantly [8, 15–17]. Since the nucleus of a cell is one of the important scatterers, these changes eventually affect the scattering property of the epithelium. The change in scattering property inevitably induces variations in the degree of polarization (DOP) [18–21]. Polarization-sensitive OCT (PS-OCT) can provide not only cross-sectional images but also information related to the polarization state of the back-scattered (or back-reflected) light from biological tissues [22–27]. In our previous study, we demonstrated that the axial profile of the DOP and the degree of circular polarization (DOCP) are proportional to the scattering property changes and can be quantified in liquid, solid, and biological samples using PS-OCT . We used Intralipid suspensions as liquid phantoms and gelatins mixed with Intralipid suspensions as solid phantoms. In liquid and solid phantoms, the axial DOP profile was fitted to a Boltzmann equation and the fitted parameter was found to be linearly dependent on the scattering changes caused by concentration changes. It was also shown that the scattering measurements using PS-OCT are considerably less sensitive to the location of the focal point as compared to the previously reported method [11, 14] that uses backreflected intensity profiles. The DOP or DOCP can also be affected by birefringence or noise. We showed that the strong near-surface signals are not affected by the noise for nonbirefringent materials . In this study, we compare our DOCP results with histological readings and test the feasibility of our technique in the diagnosis of cervical pathology.
2. Materials and Methods
2.1. PS-OCT instrument and data processing
Fig. 1 shows a schematic of a typical PS-OCT system [22–27]. It is basically a partial coherence Michelson interferometer designed to detect the polarization state of the backscattered light from tissues. We used a superluminescent diode (SLD) with a center wavelength of 1,300 nm and a full width half maximum (FWHM) of 40 nm. After transmission through a linear polarizer (LP), the horizontally polarized light is split into sample and reference arms. Quarter-wave plates (QWPs) are placed on both the sample and reference arms. The light in the reference arm passes through a QWP oriented at 22.5° with respect to the horizontal axis and is reflected by the reference mirror. After the reflected light passes through the QWP again, the light has a +45° linear polarization state. For axial scanning, we used a rapid scanning optical delay line (RSOD) method, which requires a grating and a galvanometer [29, 30]. The QWP plate in the sample arm is oriented at 45° to the horizontal direction to provide circularly polarized light to the sample. Another galvanometer and linear stage are used for lateral and transversal scanning, respectively. Light reflected from the tissue is recombined with light from the reference arm, which is then split into two orthogonal polarization states: the horizontal and vertical polarization states. The two light beams are detected by two photodiodes and digitized by a data acquisition (DAQ) board. The axial scanning rate and lateral scanning rate are 200 Hz and 1 Hz, respectively. For transversal scanning, the velocity of the linear stage is set at 0.02 mm/s. The number of data points for each A-line scan is 5,000 points. The axial and lateral resolutions are 13 µm and 15 µm, respectively.
The two orthogonal signals are digitized and then filtered by a digital bandpass filter that extracts interference signals with a Doppler frequency of 65 kHz. The amplitudes and phases are obtained from the filtered signals using Hilbert transformation. To determine the polarization state of the sample, the Stokes parameters of the light reflected from the sample are computed as follows [22–27]:
where Eox and Eoy are the amplitudes, and εox and εoy are the phases in the horizontal and vertical channels, respectively. Two of the Stokes parameters, S1 and S3, have to be exchanged because the light reflected from the sample has to pass again through the QWP oriented at 45° in the sample arm [22–24]. Then, the DOCP is calculated as .
2.2. Sample preparations
The experimental protocol followed the directions of the Institution Review Board (IRB) at the Wonju Christian Hospital, Yonsei University. Cervical conization tissues were obtained from 18 female patients (Table 1). Immediately after cone biopsy, each sample was imaged with the PS-OCT system. About 2–4 areas were imaged per sample; we obtained a total of 71 images. An image of one of the cervical samples in which the scanned areas are indicated is shown in Fig. 2. The excised samples were then stained with hematoxylin and eosin (H&E) and read by two pathologists. Later, the histological readings were compared with the PSOCT results. The PS-OCT and histology data were co-registered by marking the imaged areas using dark ink between PS-OCT and pathology readings.
We obtained backscattered images (S0) and normalized circular polarization images (S3/S0) of the cervical samples. Fig. 3 shows the typical images of tissues under various conditions. The physical dimensions of each image are approximately 2 mm (depth) × 2 mm (width). Figs. 3(a), (b), and (c) are backscattered intensity images. The corresponding normalized polarization images are shown in Figs. 3(d), (e), and (f). Figs. 3(a) and (b) are images of normal tissue samples. In Fig. 3(a), the layered structure of the cervix comprising the epithelium (e) and stroma (s) is clearly visible. However, although Fig. 3(b) was also obtained from a normal cervix, the layered structure is not visible. Fig. 3(c) was obtained from an abnormal tissue with H-SIL; the layered structure is not visible in this case as well. Therefore, it is concluded that the presence of a basement membrane, which distinguishes the epithelium from stroma, in the intensity images cannot be a good marker for precancer diagnosis. It is also shown that it is difficult to simply distinguish between the normal (Fig. 3(b)) and H-SIL (Fig. 3(c)) tissues by observing the intensity images. However, the penetration depth of a normal tissue image is deeper than that of an H-SIL tissue image due to the change in the scattering coefficient which is as expected. Figs. 3(g) and (h) are the histological images of normal and abnormal tissues obtained from the OCT-imaged areas, respectively.
Figure 4 is a 3-D volume image of a normal cervical tissue constructed from 200 2-D backscattered intensity images. The scanned area has a size of 2 mm (lateral) × 4 mm (transversal). We also constructed another 3-D volume image from 2-D polarization state images, which are not shown here. We did not average the data for the entire 3-D image. However, instead, we split each 3-D DOCP volume image into 3~9 sub-volumes and then averaged the DOCP signals for each sub-area.
The DOCP changes are caused by depolarization due to scattering, phase retardation due to tissue birefringence, noise fluctuation artifacts, etc. . In general, since the SNR of the intensity signal from the epithelium was considerably higher than 20 dB, as shown in Fig. 5(a), we excluded the possibility that the noise fluctuation significantly influenced the DOCP changes. Furthermore, we do not believe that the birefringence affected the DOCP changes in the epithelium because there are no collagen fibers in the epithelium of the cervix . On the other hand, the basement membrane comprises type IV collagen, and collagen and elastin are the major components of the stroma [9, 32]. Therefore, we hypothesize that the DOCP changes in the epithelium were induced only by the depolarization caused by scattering. Fig. 5(b) shows the axial decay profile of the averaged DOCP signal for the sample shown in Fig. 3(b). Because we were interested only in the epithelium layer, the slope of the DOCP decay was quantified from 80 µm to 250 µm, and the surface signal was excluded using the leastsquare linear fit method. Fig. 5(c) shows that the estimated slope of the DOCP signal is 0.762 mm−1. After the slope of the axial DOCP decay in the sub-volume images was averaged, we compared our results with the pathological findings.
Figure 6 is an exemplary pathological map (gray colors). Two pathologists examined all the specimens using a light microscope after the samples were cut into 12 pieces along the 1 to 12 o’clock directions. The bars in red and blue represent pathological results near the areas scanned by our PS-OCT system so that a direct comparison can be made.
In this study, since it was very difficult to accurately match the scanned areas with the pathological map, we classified a sample volume as H-SIL if at least one sub-volume was found to contain H-SIL. After all the data were collected from the 71 sample areas (3D volumes), we specifically defined a threshold value to discriminate between L-SIL and H-SIL. In our previous study, we found that the averaged slopes of the DOCP signal were 0.92 mm−1 and 1.71 mm−1 in normal and H-SIL cervix tissues, respectively . Because we had only one parameter (averaged slope of the DOCP), we could not use a classification method such as the Mahalanobis distance or a support vector machine [33, 34]. We adjusted the threshold value from 1.6 mm−1 to 2.0 mm−1 (Table 2) after setting a criterion with the averaged slope of H-SIL. The best criterion to discriminate between normal and H-SIL samples was found to be an axial decay slope of 1.8 mm−1, which produced a sensitivity and specificity of 94.7% and 71.2%, respectively (Table 3). We found one false-negative case from a sample that was blood-stained and produced a low signal. We believe that our readings can be improved if the entire cervix area can be scanned.
In Fig. 7, four more cases are displayed as additional examples. On the left-hand side of each sample, the actual picture is shown. Fig. 7(c) shows a sample that was diagnosed as LSIL from our result, but found to be H-SIL from the pathological reading.
Depolarization due to scattering is known to be a negative effect that prevents the imaging of deeper regions. However, we have used the same depolarization effect to our advantage and have suggested a new diagnostic method that uses the redundant well-known PS-OCT system. Cervical pathology has been studied using various optical imaging technologies, including conventional OCT. However, these methods measure the axial intensity decay slope from the surface; therefore, they are significantly affected by the location of the focal point. If the axial depolarization signal is used for cervix diagnosis since the axial depolarization data are normalized by the intensity data, more stable diagnoses can be performed.
Currently, we have obtained a sensitivity of 94.7% and a specificity of 71.2%. Only one false-negative case was found out of 71 samples. We find that our technique is quite sensitive in comparison to other methods. In addition, it can be further improved by implementing a full-range imaging probe.
The current trend is to combine more than one technology for specific clinical diagnosis. Intraepithelial neoplasm in the cervix is known to occur at the squamo-columnar junction; however, at times, it originates from the sub-surface region and cannot be detected by a naked-eye examination. Various surface-imaging technologies such as colposcopy and confocal microscopy have been tested for CIN diagnosis, and it would be desirable to combine both the surface and sub-surface diagnostic technologies. The PS-OCT system can be used as an alternative option for monitoring sub-surface lesions; therefore, this paper may provide a valuable diagnostic option for CIN diagnosis.
One of the problems that still limit PS-OCT applications in clinics is that it is somewhat cumbersome to use the fiber-based system. First of all, multiple input polarizations (2 or 4) have to be supplied to the system to determine the polarization state of the light reflected from the tissue, which can be problematic since longer data acquisition times are needed. In addition, the input polarization in the sample arm must always be estimated using a separate polarization analyzer since the polarization can be arbitrarily changed as the fiber probe moves. Solving these problems will increase the utility of PS-OCT in clinical diagnosis and this can be considered in future studies.
In this study, we obtained in vitro cross-sectional and polarization state images of cervical tissues by means of a PS-OCT system. When CIN, which is cervical dysplasia generated at the squamo-columnar junction, is present in the cervix, there is an increase in the nuclear size and nuclear/ cytoplasmic (N/C) area ratio. The change in the N/C area ratio, which reflects the degree of CIN, results in a change in the scattering coefficient. We quantified the slope of DOCP decay for the diagnosis of CIN. As expected, we found that the DOCP decayed faster as a function of depth for H-SIL (CIN-II, CIN-III, and CIS) due to the higher scattering. Therefore, our studies indicate that the PS-OCT system might be useful in the measurements of DOCP with depth in screening for cervical dysplasia. Generally, for elderly patients, the squamo-columnar epithelial junction sinks into the orifice of the cervix. For further studies, we will require in vivo investigations using endoscopic PS-OCT technology.
This study was supported by a grant from the Korea Health 21 R&D Project, Ministry of Health & Welfare, Republic of Korea (02-PJ3-PG6-EV07-0002). This work was also supported by the Bio-signal Analysis Technology Innovation Program (M10645010001- 06N4501-00110) of the Ministry of Science and Technology (MOST) and Korea Science and Engineering Foundation (KOSEF). This research was also supported by a grant (M103KV010024-07K2201-02410) from Brain Research Center of the 21st Century Frontier Research Program funded by the Ministry of Science and Technology, Republic of Korea.
References and links
1. A. A. Garcia, O. Hamid, and A. E1-Khoueiry, “Cervical cancer,” http://www.emedicine.com/med/topic324.htm.
2. R. A. Drezek, R. Richards-Kortum, M. A. Brewer, M. S. Feld, C. Pitris, A. Ferenczy, M. L. Faupel, and M. Follen, “Optical imaging of the cervix,” Cancer (Suppl.) 98, 2015–2027 (2003). [CrossRef]
3. I. J. Bigio, J. R. Mourant, J. Boyer, and T. M. Johnson, “Invited: Elastic scattering spectroscopy for diagnosis of tissue pathologies,” in Biomedical Optical Spectroscopy and Diagnostics, E. Sevick-Muraca and D. Benaron, eds., Vol. 3 of OSA Trends in Optics and Photonics Series (Optical Society of America, 1996), paper SP1, http://www.opticsinfobase.org/abstract.cfm?URI=BOSD-1996-SP1.
4. R. Hornung, T. H. Pham, K. A. Keefe, M. W. Berns, Y. Tadir, and B. J. Tromberg, “Quantitative nearintrared spectroscopy of cervical dysplasia in vivo,” Human Reproduction 14, 2908–2916 (1999). [CrossRef] [PubMed]
5. A. Mahadevan, M. F. Mitchell, E. Silva, S. Thomsen, and R. Richards-Kortum, “Study of the fluorescence properties of normal and neoplastic human cervical tissue,” Lasers Surg. Med. 13, 647–655 (1993). [CrossRef] [PubMed]
6. N. Ramanujam, M. F. Mitchell, A. Mahadevan, S. Thomsen, A. Malpica, T. Wright, N. Atkinson, and R. Richards-Kortum, “Development of a multivariate statistical algorithm to analyze human cervical tissue fluorescence spectra acquired in vivo,” Lasers Surg. Med. 19, 46–62 (1996). [CrossRef] [PubMed]
7. N. Ramanujam, M. F. Mitchell, A. Mahadevan-Jansen, S. L. Thomsen, G. Staerkel, A. Malpica, T. Wright, N. Atkinson, and R. Richards-Kortum, “Cervical pre-cancer detection using a multivariate statistical algorithm based on laser induced fluorescence spectra at multiple excitation wavelength,” Photochem. Photobiol. 6, 720–735 (1996). [CrossRef]
8. K.-B. Sung, R. Richards-Kortum, M. Follen, A. Malpica, C. Liang, and M. R. Descour, “Fiber optic confocal reflectance microscopy: a new real-time technique to view nuclear morphology in cervical squamous epithelium in vivo,” Opt. Express 11, 3171–3181 (2003), http://www.opticsinfobase.org/abstract.cfm?URI=oe-11-24-3171. [CrossRef] [PubMed]
9. T. Collier, M. Follen, A. Malpica, and R. Richards-Kortum, “Sources of scattering in cervical tissue: determination of the scattering coefficient by confocal microscopy,” Appl. Opt. 44, 2072–2081 (2005), http://www.opticsinfobase.org/abstract.cfm?URI=ao-44-11-2072. [CrossRef] [PubMed]
10. F. I. Feldchtein, G. V. Gelikonov, V. M. Geliknov, R. V. Kuranov, and A. M. Segeev, “Endoscopic applications of optical coherence tomography,” Opt. Express 3, 257–270 (1998), http://www.opticsinfobase.org/abstract.cfm?URI=oe-3-6-257. [CrossRef] [PubMed]
11. A. Zuluaga, M. Follen, I. Boiko, A. Malpica, and R. Richards-Kortum, “Optical coherence tomography: A pilot study of a new imaging technique for noninvasive examination of cervical tissue,” Am. J. Obstet. and Gyneco. 193, 83–88 (2005). [CrossRef]
12. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, “Optical coherence tomography,” Science 254, 1178–1181 (1991). [CrossRef] [PubMed]
13. J.-Y. Yoo, S.-W. Lee, M.-S. Kang, Y.-T. Kim, and B.-M. Kim, “Early-stage diagnosis of cervix using PSOCT,” presented at the Photonics West 2006, USA, 21–26 Jan. (2006).
14. R. O. Esenaliev, K. V. Larin, and I. V. Larina, “Noninvasive monitoring of glucose concentration with optical coherence tomography,” Opt. Lett. 26, 992–994 (2001), http://www.opticsinfobase.org/abstract.cfm?URI=ol-26-13-992. [CrossRef]
15. P. Athanassiandou, N. Kavantzas, M. Gonidi, P. Athanassiades, A. Liossi, L. Nakopoulou, E. Petrakakou, and P. Davaris, “Correlations between nuclear/cytoplasmic area ratio and classification of cervical smears,” J. Exp. Clin. Cancer Res. 20, 481–486 (2001).
16. C. R. de R. Miranda, L. Deligdisch, J. Gil, and P. Dottino, “Morphometric characterization of HPV-related lesions of uterine cervix,” J. Low. Genit. Tract. Dis. 4, 204–207 (2000).
17. D. Arifler, M. Guillaud, A. Carraro, A. Malpica, M. Follen, and R. Richards-Kortum, “Light scattering from normal and dysplastic cervical cells at different epithelial depths: finite-difference time-domain modeling with a perfectly matched layer boundary condition,” J. Biomed. Opt. 8, 484–494 (2003). [CrossRef] [PubMed]
18. A. D. Kim and M. Moscoso, “Influence of the relative refractive index on the depolarization of multiply scattered,” Phys. Rev. E. 64, 026612 (2001). [CrossRef]
20. I. Charalambous, R. Cucu, A. Dogariu, and A. Podoleanu, “Experimental investigation of circular light depolarization using polarization sensitive OCT,” Proc. of SPIE 6429, 64291S (2007). [CrossRef]
21. S. P. Morgan and M. E. Ridgway, “Polarization properties of light backscattered from a two layer scattering medium,” Opt. Express 7, 395–402 (2000), http://www.opticsinfobase.org/abstract.cfm?URI=oe-7-12-395. [CrossRef] [PubMed]
22. J. F. de Boer, S. M. Srinivas, A. Malekafzali, Z. Chen, and S. Nelson, “Imaging thermally damaged tissue by polarization sensitive optical coherence tomography,” Opt. Express 3, 212–218 (1998), http://www.opticsinfobase.org/abstract.cfm?URI=oe-3-6-212. [CrossRef] [PubMed]
23. M. G. Ducros, J. F. de Boer, H. Huang, L. C. Chao, and Z. Chen, “Polarization sensitive optical coherence tomography of the rabbit eye,” IEEE J. Select. Topics Quantum Electron. 5, 1159–1167 (1999). [CrossRef]
25. S. M. Srinivas, J. F. de Boer, H. Park, K. Keikhanzadeh, H. L. Huang, J. Zhang, W. Q. Jung, Z. Chen, and J. S. Nelson, “Determination of burn depth by polarization-sensitive optical coherence tomography,” J. Biomed. Opt. 9, 207–212 (2004). [CrossRef] [PubMed]
26. S.-W. Lee, J.-T. Oh, and B.-M. Kim, “Microscopic imaging of articular cartilage using polarizationsensitive optical coherence tomography,” J. Biomed. Eng. Res. 26, 33–38 (2005).
27. J.-T. Oh, S-W. Lee, Y.-S. Kim, K.-B. Suhr, and B.-M. Kim, “Quantification of the wound healing using polarization-sensitive optical coherence tomography,” J. Biomed. Opt. 11, 041124 (2006). [CrossRef] [PubMed]
28. S.-W. Lee, J.-H. Kang, J.-Y. Yoo, M.-S. Kang, J.-T. Oh, and B.-M. Kim, “Quantifcation of scattering changes using polarization-sensitive optical coherence tomography,” J. Biomed. Opt. (submitted). [PubMed]
29. A. M. Rollins, M. D. Kulkarni, S. Yazdanfar, R. Ung-arunyawee, and J. A. Izatt, “In vivo video optical coherence tomography,” Opt. Express 3, 219–229 (1998), http://www.opticsinfobase.org/abstract.cfm?URI=oe-3-6-219. [CrossRef] [PubMed]
30. W. K. Niblack, J. O. Schenk, and M. E. Brezinski, “Dispersion in a grating-based optical delay line for optical coherence tomography,” Appl. Opt. 42, 4115–4118 (2003), http://www.opticsinfobase.org/abstract.cfm?URI=ao-42-19-4115. [CrossRef] [PubMed]
31. T. C. Wright and A. Ferenczy, “Anatomy and histology of the cervix,” in Blaustein’s Pathology of the Female Genital Tract, R. J. Kurman, ed. (Springer-Verlag, 2002), pp. 207–224.
32. K. Carlson, I. Pavlova, T. Collier, M. Descour, M. Follen, and R. Richards-Kortum, “Confocal microscopy: Imaging cervical precancerous lesions,” Gynecologic Oncology 99, S84–S88 (2005). [CrossRef] [PubMed]
33. Y. N. Mirabal, S. K. Chang, E. N. Atkinson, A. Malpica, M. Follen, and R. Richards-Kortum, “Reflectance spectroscopy for in vivo detection of cervical precancer,” J. Biomed. Opt. 7, 587–594 (2002). [CrossRef] [PubMed]
34. G. M. Palmer, C. Zhu, T. M. Breslin, F. Xu, K. W. Gilchrist, and N. Ramanujam, “Comparison of multiexcitation fluorescence and reflectance spectroscopy for the diagnosis of breast cancer,” IEEE Trans. Biomed. Eng. 50, 1233–1242 (2003). [CrossRef] [PubMed]