Over the past few years one of the new imaging technologies we have been exploring is Optical Coherence Tomography an imaging technology that provides high definition in superficial tissue layers using infra-red. The first project, based primarily in Cleveland, attempted to develop OCT image criteria that could be used to clearly separate normal from abnormal cervical tissue. On completion of this initial experience with 50 patients we were quite confident we could differentiate between OCT images of normal cervix and OCT images of invasive cancer. OCT was able to visualize the normal microstructure of the various cervix anatomies.
In our second study, conducted primarily in the Dominican Republic (with a subpopulation from the Cleveland Clinic), 1215 OCT images with matched unmagnified visual inspection using acetic acid (VIA) and colposcope impressions and histology from 212 patients were assessed “remote from the time of the examination” by 3 reviewers. All patients had had VIA, OCT, colposcopy, and biopsies using our validation protocol. Both VIA and colposcopy showed improved specificity when combined with OCT. However, it was clear to us that “time-remote interpretation” of such a dynamic technology was sub-optimal. We have since done two real time studies in China. In Shenzhen, China 299 women were examined with the colposcope with a diagnostic impression made for each quadrant by the operating gynecologist. This was followed by OCT with real-time diagnostic impression and biopsies according to our standard micro-biopsy protocol. The result was a minimum of 4 biopsies per patient with matched OCT images and real-time impressions, plus matched digitized colposcopy images and colposcopy-based diagnoses from each patient. This was our first clinical investigation using the OCT technology “real time”. In addition to improving specificity we have identified two measurable OCT image characteristics that appear to be correlated with the evolving pathologic disease process.
In April of 2008 we completed the fourth of our planned clinical investigations, this time in the “Buyi-Miao Autonomous District of Guizhou Province China in collaboration with the Peking University Shenzhen Hospital. This study was designed to explore potential application of OCT technology in low-resource settings. Our primary objective was to determine the sensitivity, specificity, positive and negative predictive values for “real time” OCT just using the naked eye (VIA) to guide the OCT device. Our secondary objective was to examine the potential for OCT to improve the sensitivity and specificity of VIA and the ease of use of OCT in a real time clinical setting in a rural low resource environment. The subjects came from the villages surrounding the four main towns: Wengan, Sandu, Libo and Duyun, all in Guizhou Province, China. During the Guizhou trial there were several OCT observations noted, such as the “epithelial brightness” first discovered in our Shenzhen studies. However, we found ourselves confused by the multiple variables we had identified and we were not able to adequately incorporate the change in brightness into our real-time “visual” interpretation. In fact we now believe that we have identified more image characteristics than we can possibly evaluate in real-time, without the help of a mathematical algorithm and “computer aided diagnosis.”
Statistical Significance Achieved for Epithelial Brightness as a Defining Characteristic to Evaluate Pre-invasive and Invasive Cervical Cancer
Observations made during our two China OCT studies led us to analyze the differences in the brightness of the cervical epithelium on Optical Coherence Tomography (OCT) images as a potential distinguishing characteristic of normal, low-grade, high-grade, and cancer histologies. The data from the 300 women who participated in a real-time study of OCT as a diagnostic adjunct to colposcopy and 183 women who participated in a real-time study of OCT as a diagnostic adjunct to VIA were combined to compare the relationship between the brightness of the OCT images to corresponding histology. The abnormal images were expressed in decibels and weighted by the number of images per location. The standard deviation of the average difference in brightness by all patients was also measured. To generate the brightness unit for analysis, the “normal” reading was subtracted from an abnormal reading calculating a change in brightness.For each histological grade mean difference in brightness from normal was calculated. . All brightness measures were a log scale. Mean brightness was 0.16, 1.56, 3.36, and 4.71 for squamous metaplasia, CIN II, CIN III, and cancer respectively. Mean brightness differed significantly between each histological grade (p-values 0.000) for the comparisons of CIN II to CIN III, CIN II to cancer, and squamous metaplasia to cancer. For the comparison of mean brightness for CIN III to cancer p=.008. We conclude that epithelial brightness is an important component to include in the development of mathematical algorithms to use for the diagnostic interpretation of OCT generated images of the uterine cervix.