Optical coherence tomography (OCT) is similar to ultrasound pulse-echo imaging, with optical rather than acoustic reflectivity being measured. The OCT is a harmless infrared light technology that uses low-coherence interferometry to produce an image based on optical scattering from an internal microstructure. Compared with existing imaging methods, image resolution is superior at lesser tissue depths and gives structural information of the tissues exposed.1 The compact imaging consists of an imaging console, a keyboard/touch pad for data entry, and a detachable fiber-optic probe. The imaging console contains optical and electrical components, including the light source that provides near infrared light, 980 to 1320 nm in wavelength, which is directed from the scanner through the 2.7-mm diameter probe toward the patient's tissue. The optical light is backscattered from the patient's tissue, collected by the probe's fiber, and combined with an internal reference signal to produce a high-spatial resolution image of the superficial tissue microstructure with a 10-μm in-depth and 25-μm lateral resolution acquiring 200 × 200 images in 1.5 seconds.
The earliest publication of OCT and its potential clinical impact was available in 1991 by Huang.1 This technology's clinical value was first recognized in ophthalmology at the New England Eye Center and Tufts University in studies of more than 5000 patients in which OCT proved to be useful in detecting early disease progression for various retinal diseases.2-4 Various research groups from around the world, including The Cleveland Clinic, are currently studying the use of OCT in a variety of areas including gastrointestinal, pulmonary, urinary, gynecologic, dermatologic, dental, and cardiovascular settings.5-11
OCT may provide a major technical advance for imaging of the cervix. It has the potential to be a true "optical biopsy." Moreover, it has the potential to be another widely accessible noninvasive tool in the arsenal of cancer detection and prevention. In a pilot study done at The Cleveland Clinic Foundation, OCT accurately characterized normal squamous epithelium of the cervix (100%) and histologically proven cervical intraepithelial neoplasia 2 to 3 (CIN 2-3) (89%).12 These facts give the technology potential ability of providing point-of-service diagnosis and on-site management, if diagnostically comparable to a cervical biopsy.
In the developing world where screening methods requiring minimal infrastructure are most desirable, using OCT to potentially eliminate biopsy and allow point-of-service diagnosis and treatment with increased specificity would be ideal. An algorithm using a technique such as unaided visual inspection with acetic acid (VIA), which has a sensitivity for high-grade preinvasive disease and cancer greater than a conventional Papanicolaou test but a poorer specificity, could potentially be significantly improved by using OCT as a secondary screen.
In the largest prospective trial of OCT images matched to cervical biopsies, investigators from The Cleveland Clinic Foundation, Hospital Maternidad Nuestra Senora de la Altagracia at Dominican Republic, and Preventive Oncology International assessed OCT sensitivity and specificity alone and as an adjunct to VIA and colposcopy in the diagnosis of preinvasive and invasive lesions of the uterine cervix.8 They reported OCT as a fair diagnostic adjunct (receiver operating characteristic curve, 0.73) to VIA and colposcopy in accurately identifying lesions distant from the squamocolumnar junction (SCJ). The poor specificity (∼50%-84%) reported in that study, the investigators believed, was primarily the result of the non-real-time interpretation of the images, creating isolation from clinical input and probable image-biopsy mismatch at the SCJ.
A subsequent study completed from November to December 2007 in Shenzhen, China, was an evaluation of the sensitivity and specificity of real-time OCT as an adjunct to colposcopy (Z. Liu et al, unpublished data). This study showed that when OCT was added to colposcopy, specificity increased from 0.83 to 0.93. With the additional 1237 OCT images obtained, image characteristics with the potential to be added to our diagnostic criteria were evaluated, and plans were made to incorporate them into the current study. For example, rather than using pure pattern recognition, brightness of signal intensity would be added to our diagnostic criteria.
The current study was designed to explore potential applications of OCT technology in low-resource settings. Here, we report the sensitivity and specificity values for OCT guided by VIA in the detection of lesions equal to or greater than CIN II in a real-time clinical evaluation, in Guizhou, China. We discuss the ease of use of OCT in a real-time clinical setting and the potential for OCT to improve the sensitivity and specificity of VIA.
PATIENTS AND METHODS
The institutional review boards for human subject research of both the Cleveland Clinic Foundation and the Peking University Shenzhen Hospital in Shenzhen, China, approved the study. This clinical study was a prospective cross-sectional comparative trial coordinated by the Renmin Hospital in the Buyi-Miao Autonomous District of Guizhou Province, China, in collaboration with the Peking University Shenzhen Hospital and Preventive Oncology International, Inc, affiliated with the Cleveland Clinic. To be qualified to enter the study, participants had to meet the following inclusion criteria: (a) nonpregnant female between the ages of 30 and 50 years; (b) had to sign an informed consent document stating that they understood the investigational nature of the proposed study; and (c) physically and mentally willing to comply with all study requirements, especially the conduct of a colposcopy examination. Exclusion criteria included: (a) inability to provide adequate informed consent and/or comply with the study requirements; (b) pregnancy; (c) prisoner status; (d) a prior hysterectomy; or (e) receiving prior treatment for a preinvasive or invasive cervical cancer. Dr Huang Yuxia, the chief of gynecology at Renmin Hospital, coordinated the recruitment of patients with doctors from local hospitals until the goal of 1000 participants was reached. Study participants came from the surrounding villages of 4 main towns: Wengan, Sandu, Libo, and Duyun, all in Guizhou Province, China. Informed consent was obtained immediately preceding the initial screening during which a physician obtained a human papillomavirus (HPV) test and liquid-based cervical cytology. All screening was performed by gynecologists from the Peking University Shenzhen Hospital. OCT was performed and interpreted by one of the authors (J.L.B.), and the VIA was performed and interpreted by one of the authors (N.R.).
Women with abnormal cervical cytology or positive HPV test results were requested to return to the clinic for VIA, OCT imaging, colposcopy, and cervical biopsies. After receiving a thorough explanation of the study, participants who continued to meet the inclusion criteria were enrolled in the study. First, participants were examined using VIA to assess the cervix for abnormalities. The cervix was divided into quadrants for examination: quadrant I (12:00-3:00), quadrant II (3:00-6:00), quadrant III (6:00-9:00), and quadrant IV (9:00-12:00). All cervical quadrants were assessed, and the diagnoses were recorded by quadrant. The VIA diagnoses were: N (normal), L (low grade), H (high grade), or C (cancer). A normal cervix had no white lesions. A diagnosis of low grade showed pale white lesions that might or might not abut the SCJ. Areas defined as high grade had dense white lesions with sharp borders; 1 border of these high-grade lesions always abutted or was close to the SCJ. Cancer was diagnosed when a friable mass with an irregular surface was seen.
Second, the cervix was examined with OCT by quadrant guided by the VIA. The OCT images were graded using our previously described scale (Escobar 2006): normal if a well-organized simple 2-layer structure was seen with a sharp interface between the surface (squamous) epithelium and the underlying stromal layer (dense connective tissue); abnormal if the tissue was unstructured with no apparent interface present; and indeterminate if irregularities on the images suggested artifacts or physiological alterations (swelling of the epithelium, edema of the stromal layer, or inflammation) and did not meet diagnostic criteria for normal or abnormal. When the OCT image was read as either indeterminate or abnormal, efforts were made to compare the epithelial brightness of the quadrant's image with that of the baseline image to see if any information could be gleamed from the brightness of the epithelium (as viewed by the human eye) while at the bedside. If a quadrant was abnormal on VIA examination (≥low grade), OCT readings were taken in the areas of VIA abnormality. In normal quadrants, OCT readings were obtained at 2, 4, 8, or 10 o'clock depending on the quadrant. Baseline OCT readings were obtained at 12 and 3 o'clock in the transformation zone to use for diagnostic comparison. All OCT images were read at the time of image capture by the operator. The OCT diagnoses were as follows: "0" (normal), "1" (indeterminate), or "2" (abnormal). Third, an assessment by colposcopy and biopsies obtained with a 2-mm bronchoscopy forceps based on the Preventive Oncology International standard biopsy protocol.13 The cervix was examined by quadrant. All colposcopically detected abnormalities were recorded and biopsied by quadrant. In normal quadrants, biopsies were obtained at the SCJ at 2, 4, 8, and 10 o'clock depending on the quadrant. An endocervical curettage was performed in all patients. All patients therefore had a minimum of 5 biopsies: SCJ 2, 4, 8, and 10 o'clock and an endocervical curettage.
A team of pathologists read the pathology, and 1 gynecologic pathologist served as the final reference and quality control (QC). All women found to have high-grade precancer or true invasive cancer were cared for according to standard treatment protocols at the local institution.
A data technician reviewed all data forms and other documents to ensure completeness and for QC before the patient left the clinic. Once the QC was completed, the data were entered in duplicate into the computer database. The final database was reconciled at the P.O.I. epidemiology and biostatistical center at Northwestern University.
The performance characteristics of VIA and VIA-directed OCT were evaluated by calculating the sensitivity and specificity according to the standard definitions. Because all eligible women received the reference standard (cervical biopsy), the previously mentioned estimates were directly calculated. Women in this study had a minimum of 1 matched set of VIA, OCT, and biopsy results for a given lesion per quadrant, and a minimum of 4 sets overall (four quadrants). Generalized estimating equations were used to generate estimates of sensitivity and specificity with 95% confidence intervals (CIs) using all matched sets, making the level of analysis of the lesion, while accounting for the additional correlation from using multiple records per woman. In the per-woman analysis, the VIA, OCT, and biopsy results were collapsed across quadrants, and the highest value for each represented the value for that woman.
A third analysis was completed that used 1 matched set per woman. The matched set was chosen by locating the worst biopsy for that woman and analyzing it with its paired OCT and VIA. Logistic regression analysis was used to generate sensitivity and specificity estimates with corresponding 95% CIs. Test performance was determined both for greater than or equal to CIN 2 and greater than or equal to CIN 3. All data analyses were performed using STATA 9 (StataCorp LP, College Station, TX).
Of the 1000 participants initially screened, 175 (17.5%) were HPV positive, 93 (9.3%) had abnormal cervical cytology that was greater than or equal to atypical squamous cells of undetermined significance, and 211 (21.1%) were either HPV positive or had abnormal cervical cytology. The VIA, OCT, colposcopy, and biopsies were completed on 182 (86.7%) of 211 women who are included in the analysis. Twenty-one women were lost to follow-up, and an additional 8 returned when OCT was no longer available.
Table 1 shows the demographic characteristics of the 182 women who were included in these analyses. The median age was 41 years, the median number of pregnancies was 3, and the median number of live births was 1. Pathology of the cervical biopsies revealed that 15% (27) of the participants had lesions that were greater than or equal to CIN 2, whereas 8% (15) had lesions greater than or equal to CIN 3. One participant's cancer diagnosis came from the endocervical curettage, which does not have a matched OCT, therefore, 26 participants with 39 greater than or equal to CIN 2 lesions and 14 participants with 23 greater than or equal to CIN 3 lesions were included in the by-lesion and by-women and lesion analyses.
For VIA alone, using high-grade as positive, the sensitivity and specificity in detecting lesions greater than or equal to CIN 2 was 43% (95% CI, 28-60) and 96% (95% CI, 95-98), respectively, when analyzing data by lesion (Table 2). With the addition of OCT, the sensitivity increases to 62% (95% CI, 45-78), with a specificity of 80% (95% CI, 76-83), also analyzed by lesion (Table 3).
If a per-woman analysis is performed, such that the worst VIA, OCT, and biopsy results in any quadrant are used, the sensitivity and specificity for VIA are similar: 33% (95% CI, 17-54) and 93% (95% CI, 88-96), respectively, in detecting lesions greater than or equal to CIN 2. The sensitivity and specificity for VIA-directed OCT in a per-woman analysis is 59% (95% CI, 39-78) and 65% (95% CI, 57-73), respectively.
Finally, if an analysis is performed such that the worst biopsy and its corresponding results are selected in each woman, the sensitivity and specificity for VIA are 31% and 96%, respectively, and when OCT is added to VIA, the sensitivity increases to 50%, whereas the specificity decreases to 80%. These results are summarized in Tables 4 and 5. Using VIA to direct OCT provides a large gain in sensitivity, but the loss in specificity is equally large.
The current study was designed to explore the potential application of OCT technology in low-resource settings using unmagnified VIA to guide OCT. The VIA has previously performed as more sensitive for the detection of precancerous lesions than the published data on conventional cytology but with a lower specificity, resulting in more false-positives.13,14 In low-resource settings, the addition of OCT to VIA, we hoped, could improve the detection of lesions greater than or equal to CIN 2, such that a highly specific diagnosis could be made, and treatment could be performed all within the same visit. Primary VIA models are currently in place in a number of countries with patients being treated after VIA screening with cryotherapy.15,16 However, such a model, although reasonable in some settings, lends itself to overtreatment of patients who have no more than nonneoplastic HPV infections that have a high probability of clearance. We sought to determine if the application of OCT technology may be of benefit in such models and perform as an "optical biopsy" with high specificity. The optical biopsy could be taken from areas of the cervix that appear suspicious by VIA, potentially increasing the true-positives. Furthermore, because OCT can be performed in a real-time setting, it offers information that can be used within a single-episode point-of-care protocol.
Unfortunately, after a steady increase in our diagnostic accuracy parameters from our prior studies, this particular trial provided us with surprising results. Not only did VIA perform extremely well in terms of specificity in the HPV-positive population, but OCT influenced the specificity negatively.
The primary objective of this study was to determine the sensitivity and specificity of OCT added to VIA in the detection of lesions equal to or greater than CIN 2 in a real-time clinical evaluation. With the addition of OCT, the sensitivity of VIA increased in all analyses (by lesion, by woman, by woman/biopsy) for the detection of greater than or equal to CIN 2, and the detection of greater than or equal to CIN 3 when using high grade as the cutoff but decreased when detecting greater than or equal to CIN 3 using low grade as positive. However, with the increase in sensitivity, there was a loss in specificity. One reason for the loss in specificity may be because the specificity with VIA alone was high (63%-96%), in many cases leaving little room for improvement. In this study, HPV screening was used as a primary screen, which is known to be very sensitive, much more so than cytological screening alone.13,17 Our patient pool was thus enriched with true-positives. From this study, we are able to conclude that the addition of OCT to VIA when VIA is used as a secondary screen with primary HPV screening increased sensitivity but not specificity. We wonder how OCT would work if VIA was used as a primary screen.
We acknowledge that the sensitivity of VIA in this study is lower than what has been reported by others.13,18-23 The sensitivity for VIA in several published cross-sectional studies have varied from 29% to 95% compared with 31% to 93% in our study. Our study uses a very stringent biopsy protocol in which at least 5 biopsies were performed on every woman. With such a biopsy protocol, one is more likely to detect microscopic disease that is not visible by VIA or colposcopy, thus contributing to a lower sensitivity in these screening modalities.24
The secondary objective was to examine the ease of use of OCT in a real-time clinical setting and the potential for OCT to improve the sensitivity and specificity of VIA. The OCT was in fact very easy to use, requiring an additional 1 to 3 minutes to screen all 4 quadrants. However, OCT technology is rater dependent because it is based on visual interpretation and therefore subject to greater variability than an objective modality.11 Abnormal areas near the SCJ are often the most severe and therefore very important diagnostically. However, because of the unstructured appearance of columnar epithelium, this region is often the most difficult to interpret.
Very critical to the results in this trial was our effort to incorporate new diagnostic criteria we believed would enhance our diagnostic acumen. However, we were unable to visually discriminate those elements that our computer analysis has since revealed to us as indeed significant differences in histopathology (Belinson et al., unpublished data).
The series of OCT studies that have been performed all point to a potential role for OCT in cervical cancer screening, but additional studies and advances are needed. Although our previous study in Shenzhen, China, helped establish new diagnostic criteria, taking into account signal intensity and relative location within the squamocolumnar/endocervical region, further research within this field is needed.
A number of projects are in process and planned for the future; we believe these will advance the role of OCT in gynecology. We will use currently available technology to decrease scan time used in our studies to date (0.67 frames per second) to 8 frames per second and eventually to 30 frames per second. This should enhance the surveillance of the SCJ (a critical region in cervical carcinogenesis). We will develop a larger-diameter fiber-optic probe, allowing for visualization of the endocervix, SCJ, and transformation zone within a single frame. In addition, we will design an ergonomic disposable sterile probe sheath designed for use by the medical practitioner for the lower genital tract. We also hope to develop an inexpensive OCT system designed for low-resource applications. Finally, if our recent experience holds true, the real potential of this exciting technology will be realized when a computer algorithm is generated to aid in image interpretation.
1. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science
2. Guedes V, Schuman JS, Hertzmark E, et al. Optical coherence tomography measurement of macular and nerve fiber layer thickness in normal and glaucomatous human eyes. Ophthalmology
3. Wilkins JR, Puliafito CA, Hee MR, et al. Characterization of epiretinal membranes using optical coherence tomography. Ophthalmology
4. Hee MR, Baumal CR, Puliafito CA, et al. Optical coherence tomography of age-related macular degeneration and choroidal neovascularization. Ophthalmology
5. Chen Y, Aguirre AD, Hsiung PL, et al. Effects of axial resolution improvement on optical coherence tomography (OCT) imaging of gastrointestinal tissues. Opt Express
6. Tsuboi M, Hayashi A, Ikeda N, et al. Optical coherence tomography in the diagnosis of bronchial lesions. Lung Cancer
7. Manyak MJ, Gladkova ND, Schwartz A, et al. Evaluation of superficial bladder transitional-cell carcinoma by optical coherence tomography. J Endourol
8. Escobar PF, Rojas-Espaillat L, Tisci S, et al. Optical coherence tomography as a diagnostic aid to visual inspection and colposcopy for preinvasive and invasive cancer of the uterine cervix. Int J Gynecol Cancer
9. Gambichler T, Regeniter P, Bechara FG, et al. Characterization of benign and malignant melanocytic skin lesions using optical coherence tomography in vivo. J Am Acad Dermatol
10. Shemesh H, van Soest G, Wu MK, et al. The ability of optical coherence tomography to characterize the root canal walls. J Endod
11. Chia S, Raffel OC, Takano M, et al. Association of statin therapy with reduced coronary plaque rupture: an optical coherence tomography study. Coron Artery Dis
12. Escobar PF, Belinson JL, White A, et al. Diagnostic efficacy of optical coherence tomography in the management of preinvasive and invasive cancer of uterine cervix and vulva. Int J Gynecol Cancer
13. Belinson J, Qiao YL, Pretorius R, et al. Shanxi Province Cervical Cancer Screening Study: a cross-sectional comparative trial of multiple techniques to detect cervical neoplasia. Gynecol Oncol
14. McCrory DC, Matchar DB, Bastian L, et al. Evaluation of Cervical Cytology
. Evidence Report/Technology Assessment No. 5
; AHCPR Publ no 99-E010. Rockville, MD: Agency for Health Care Policy and Research; 1999.
15. Denny L, Kuhn L, De Souza M, et al. Screen-and-treat approaches for cervical cancer prevention in low-resource settings: a randomized controlled trial. JAMA
16. Sankaranarayanan R, Rajkumar R, Esmy PO, et al. Effectiveness, safety and acceptability of "see and treat" with cryotherapy by nurses in a cervical screening study in India. Br J Cancer
17. Belinson JL, Qiao YL, Pretorius RG, et al. Shanxi Province cervical cancer screening study II: self-sampling for high-risk human papillomavirus compared to direct sampling for human papillomavirus and liquid based cervical cytology. Int J Gynecol Cancer
18. Megevand E, Denny L, Dehaeck K, et al. Acetic acid visualization of the cervix: an alternative to cytologic screening. Obstet Gynecol
19. Sankaranarayanan R, Wesley R, Somanathan T, et al. Visual inspection of the uterine cervix after the application of acetic acid in the detection of cervical carcinoma and its precursors. Cancer
20. Sankaranarayanan R, Shyamalakumary B, Wesley R, et al. Visual inspection with acetic acid in the early detection of cervical cancer and precursors. Int J Cancer
21. Visual inspection with acetic acid for cervical-cancer screening: test qualities in a primary-care setting. University of Zimbabwe/JHPIEGO Cervical Cancer Project. Lancet
22. Sankaranarayanan R, Wesley R, Thara S, et al. Test characteristics of visual inspection with 4% acetic acid (VIA) and Lugol's iodine (VILI) in cervical cancer screening in Kerala, India. Int J Cancer
23. Qiao YL, Sellors JW, Eder PS, et al. A new HPV-DNA test for cervical-cancer screening in developing regions: a cross-sectional study of clinical accuracy in rural China. Lancet Oncol
24. Yang B, Pretorius RG, Belinson JL, et al. False negative colposcopy is associated with thinner cervical intraepithelial neoplasia
2 and 3. Gynecol Oncol
Keywords:Copyright © 2010 by IGCS and ESGO
Cervical intraepithelial neoplasia; Optical coherence tomography (OCT); Visual inspection with acetic acid (VIA)