PWE-087 Development of an auto-fluorescence index to improve the efficacy of auto-fluorescence endoscopy in detection of early neoplasia in patients with Barrett's oesophagus

Abstract

<h3>Introduction</h3> Auto-fluorescence imaging (AFI) endoscopy is a novel imaging technique that can identify neoplastic tissue by highlighting differences in tissue fluorescence properties and detect early neoplasia not detectable by conventional white light endoscopy. Normal areas appear “green” and abnormal areas are highlighted as “red” when using AFI endoscopy. Although it has a high sensitivity, it is associated with high false positive rates up to 80%.1 <h3>Methods</h3> To determine if auto-fluorescence index, defined as the ratio of the intensity of red and green tone between normal and abnormal AFI areas in patients with Barrett9s oesophagus (BO), can improve the specificity and accuracy of AFI endoscopy. Images of patients with both AFI true and false positive areas of BO were selected from two prospectively collected data sets using the Olympus Lucera-Hyperpro video endoscopy systems and GIF Q240FZ and GIFQ260FZ gastroscopes and used as training and test data sets, respectively. Biopsies from these areas had been confirmed by two specialist GI pathologists to have no dysplasia or high-grade dysplasia/carcinoma. Lesions with indefinite or low grade dysplasia were excluded. The images were de-gamma corrected, split into component channels and the mean intensity of the red and green tone at the abnormal area and background (normal areas) were quantified blinded using Image J software (NIH). Statistical analysis was done using SPSS V.17 (SPSS Inc). An optimal cut off was developed from the training set and applied blinded to the test set in order to determine the usefulness of the AFI index. <h3>Results</h3> There were 82 images from of AFI abnormalities in the training set. Of these 19 had high-grade dysplasia (HGD), 18 had carcinoma (CA), 38 had intestinal metaplasia (IM) and seven were from areas of gastric mucosa. The test set comprised of 40 images from 27 patients of which 24 had IM, six had HGD and 10 had CA. The area under the receiver operating curve (AUROC) for the auto-fluorescence index ratio (AFI ratio) in the training set was 0.9 (SE=0.3) with an optimal cut off of 1.3. Using an AFI ratio of 1.3 on the 40 images in the test set only one AF area with IM and 1 area with CA were misclassified. The sensitivity of AFI in the test set was 94% (95% CI 80 to 98), specificity was 96% (95% CI 86 to 99), the likelihood ratio of a positive test was 22.5 (95% CI 5.8 to 85). Using the AFI index reduces the false positive rate in the test set from 60% to 6%. <h3>Conclusion</h3> The AFI ratio shows promise as a tool to improve the diagnostic capabilities of AFI endoscopy. AFI index of &gt;1.3 resulted in a 10-fold reduction in the false positive rate of AFI endoscopy.