Narrow-band imaging in the evaluation of villous morphology: a feasibility study assessing a simplified classification and observer agreement

Abstract

To determine the utility of narrow-band imaging with optical magnification (NBI-Z) in the evaluation of villous morphology. Patients considered at risk of having celiac disease were invited. After standard endoscopy, they underwent further evaluation with NBI-Z which was digitally recorded. Targeted biopsies of each area videoed with NBI-Z were performed and tissue sent for histopathological analysis. Two expert endoscopists then selected the best representative videos (developmental phase). Next, 41 representative images of these videos were classified as follows: villous patterns present (N) or absent (A). Images showing absence of villi were then classified as cerebriform (C) or flat (F), corresponding to partial or total villous atrophy respectively. Three NBI-Z-naïve endoscopists then graded the videos. They underwent an interactive training session (learning phase) with video and images from a digital library before embarking on the actual assessment. To test for reproducibility, all videos were randomly reordered and graded again after a week. Forty-one videos (10 celiac disease, 31 normal) from 21 patients (3 celiac disease, 18 normal) were analyzed. The overall sensitivity and specificity in correctly distinguishing the presence or absence of villi were 93.3% and 97.8% respectively, with interobserver and intraobserver agreement (kappa, κ) at 0.82 and 0.86. The sensitivity and specificity in differentiating partial from total villous atrophy were 83.3% and 100%, κ at 0.73 and 0.68 respectively. Using a simplified classification, we demonstrated the feasibility of using NBI-Z to detect villous atrophy in patients presenting with suspected celiac disease.