Linked Color Imaging and Color Analytic Model Based on Pixel Brightness for Diagnosing H. Pylori Infection in Gastric Antrum

Abstract

Objectives: Linked color imaging (LCI) helps to differentiate minor mucosal changes, which can be objectively judged by red–green–blue pixel brightness. However, whether this color analytic model based on pixel brightness can be applied to diagnose Helicobacter pylori infection remains unknown. Methods: Consecutive adult patients with indications and underwent esophagogastroduodenoscopy for the 1 st time were enrolled in the training (n=166) and validation (n=79) set. Demographic and clinical characteristics were recorded. Target region in gastric antrum was pictured before biopsy for rapid urea test, and pixel brightness was calculated by MATLAB software. Results: In training set, 25 patients had H. pylori infection. Pixel brightness for R and B in patients with H. pylori infection was greatly higher than those in patients without H. pylori infection (R: 210.203±27.233 vs. 196.401±29.018, p=0.043; B: 127.621±26.112 vs. 125.334±27.812, p=0.025). At the cut off of R = 210 and B = 127, the specificity and sensitivity were 0.696 and 0.701. In validation set, 10 patients had H. pylori infection and the findings were consistent with those in training set. Conclusion: Color analytic model based on pixel brightness under LCI was useful in diagnosing H. pylori infection in gastric antrum.