According to the Montreal consensus, Barrett’s esophagus is a complication of GERD and is characterized by a change of normal squamous epithelium to the columnar epithelium and the development of intestinal metaplasia. Barrett’s esophagus is a known precursor for esophageal adenocarcinoma.

OBJECTIVE

We described a computer-assisted diagnostic algorithm for patients with Barrett’s esophagus based on narrow-band imaging (NBI) and magnifying endoscopy.

MATERIAL AND METHODS

69 people with Barrett’s esophagus were included in the study. 48 people from study group were examined by NBI and magnifying endoscopy; 21 people from control group were examined by conventional white-light endoscopy. Biopsy from metaplastic epithelium segments was sent for histology. The developed diagnostic algorithm based on statistical methods (Mann-Whitney test and Spearman’s rank correlation test) was used to analyze data. We used a neural network to objectify and automate the diagnosis.

RESULTS

Atrophy, intestinal metaplasia, and reticular sclerosis were more often detected in the main group. Mucosal characteristics such as irregular pattern, subepithelial angioarchitectonics, dysplasia, intestinal metaplasia, and gastric metaplasia were the most informative precursors of neoplasia. The algorithm validation accuracy using neural networks was 84.15.

CONCLUSION

NBI with magnifying endoscopy has an advantage over the conventional white-light endoscopy for diagnosing precancerous changes in esophageal mucosa. The proposed diagnostic algorithm will help improve the diagnosis of precancerous conditions in patients with Barrett’s esophagus.