Quantitative diagnostic algorithm using endocytoscopy for superficial nonampullary duodenal epithelial tumors

Background and AimOptical biopsy using endocytoscopy for superficial nonampullary duodenal epithelial tumors (SNADETs) is practical; however, a diagnostic algorithm has not been established. The aim of this study was to determine correlations of endocytoscopic findings of SNADETs with histology using computer analysis and to establish an algorithm. MethodsEndocytoscopic images and histological images of duodenal lesions from 70 patients were retrospectively collected. The numbers of glands and densely stained areas with methylene blue (DSMs) per 1 mm(2) and the percentage of DSMs per screen in endocytoscopy were determined. Moreover, correlations in DSMs and glands between endocytoscopy and histological images were analyzed. Histopathological diagnoses were assessed according to the revised Vienna classification. The primary outcome was correlation between the number of glands in endocytoscopy and that in histology. Finally, a diagnostic algorithm for endoscopic intervention of SNADETs with a statistical program command was established. ResultsThe number of glands in endocytoscopic images was correlated with that in histopathological images (rho 0.64, P < 0.001). There were significant differences in the mean number of glands between category 4/5 and category 3 (P = 0.03) and the mean percentage of DSMs between category 4/5 and category 1 (P < 0.001). When the cutoffs for the number of glands and percentage of DSMs were set at 47 per 1 mm(2) and 20.8% in one screen, respectively, the area under the ROC curve was 0.89. ConclusionsEndocytoscopic images of SNADETs reflect histopathological atypia, and computer analysis provides a practical diagnostic algorithm for endoscopic intervention.