This study demonstrates that OCT can be used to assess esophageal epithelial surface maturation and gland architecture to accurately diagnose IMC and HGD in patients with BE. This study defines OCT image criteria for dysplasia and systematically assesses the accuracy of endoscopic OCT to identify dysplastic tissue in subjects with BE. Although each of the OCT image features correlated positively with a diagnosis of IMC/HGD, the dysplasia index most accurately diagnosed IMC/HGD.
Histopathology is considered the gold standard for tissue diagnosis but is known to be imperfect. For a histopathologic diagnosis of IMC/HGD in BE, intraobserver and interobserver agreement is only moderately good (κ = 0.65 and 0.43, respectively).18
Thus, this difficulty in rendering a diagnosis of HGD must be taken into account when comparing the results of a noninvasive optical test with the gold standard of histopathology. In our study, all biopsy specimens were independently reviewed by 2 pathologists, and a consensus opinion was rendered in those cases in which there was initial diagnostic disagreement. The interobserver agreement in this study was excellent (κ = 0.89) and provides confidence in the results because the histopathologic diagnoses were rendered with consistency. Great care was taken to coordinate our OCT imaging sites with the biopsy locations. However, the precision of our registration method is on the order of several millimeters, and because IMC/HGD can be a focal process,22,23
it is possible that some of the misdiagnoses by OCT were due to poor registration of imaging and biopsy sites.
Many of the subjects in this study were undergoing endoscopy before consideration of photodynamic therapy or other therapy and therefore had known diagnoses of IMC/HGD. This accounts for the high prevalence of IMC/HGD in our study sample compared with the general population of patients with BE undergoing surveillance. Because IMC and HGD require similar management strategies and because the histopathologic distinction between the 2 diagnoses can be problematic,20
IMC diagnoses in this study were combined with HGD cases into one group representing advanced dysplasia.5
A subgroup analysis of OCT features of IMC and HGD was not conducted because of the small numbers involved. Further study examining OCT features of HGD versus those of IMC within a larger sample set might be warranted.
The investigator who analyzed the OCT images (G.J.T.) is trained in histopathology and has extensive experience with OCT imaging. This familiarity and expertise were necessary to apply the histopathology criteria to OCT images in this research setting but might have resulted in greater agreement between OCT and histopathology than would be obtained by a reader with less experience. No studies have addressed the interobserver agreement of OCT diagnoses in esophageal imaging, and it has not been established whether these results can be replicated in a general population of potential OCT image readers. Thus, although this proof of principle study suggests that histopathologic criteria can be used to grade dysplasia in OCT images of BE with good accuracy, our results require further investigation and replication on a broader scale. Furthermore, it might be possible in the future for computer software algorithms to be developed to render these diagnoses in a consistent and objective manner.
Approximately 27% of the 242 images were discarded as a result of inadequate image quality. Of these discarded images, 30% corresponded to a diagnosis of IMC/HGD. This proportion is similar to the proportion of IMC/HGD images reviewed (27%). Therefore, images of IMC/HGD were not disproportionately removed from the data set.
Pfau et al24
demonstrated the diagnostic potential of real-time visual assessment of OCT light reflectivity in colon tubular adenomas as a model of dysplasia. Our study provides further evidence that OCT imaging can accurately identify epithelial dysplasia in the lumen of the gastrointestinal tract. In addition, our results compare favorably to other imaging modalities in the detection of IMC/HGD in BE,25–27
but the method differs in that our approach uses accepted histopathologic diagnostic criteria. Although the current resolution of OCT does not allow direct discrimination of cellular and nuclear features associated with dysplasia, architectural abnormalities are also associated with progression of this disease. Our results indicate that identification of these architectural changes, which are well within the resolution of current OCT technology, can provide accurate discrimination of IMC/HGD in esophageal specialized intestinal metaplasia.
OCT technology is advancing rapidly. Some currently available OCT systems include Doppler and birefringence-sensitive imaging,28
which might provide additional contrast for improved diagnoses of IMC/HGD. Higher resolution OCT systems have also been demonstrated.29
However, before OCT can be used as a routine surveillance modality in BE, the system to deliver the OCT catheter to the esophageal lining will need to be simple, comfortable to the patient, and applicable in a large population. If the results of this study can be confirmed in a controlled, prospective analysis and combined with technologic improvements to make the device more user and patient friendly, OCT has the potential to improve the surveillance of BE.