TY - JOUR
AU1 - Guimarães, Pedro
AU2 - Keller, Andreas
AU3 - Fehlmann, Tobias
AU4 - Lammert, Frank
AU5 - Casper, Markus
AB - Endoscopy news Deep- learning based detection of gastric precancerous conditions 1 1 1 2 Pedro Guimarães, Andreas Keller, Tobias Fehlmann, Frank Lammert, Markus Casper ► Additional material is MEssagE Olympus scopes (GIF-Q160, GIF- Q160Z, GIF- published online only. To view Conventional white- light endoscopy has high 1TQ160, GIF- Q165, GIF- H180, GIF- H190; please visit the journal online interobserver variability for the diagnosis of gastric Olympus Europe, Hamburg, Germany). Images (http:// dx. doi. org/ 10. 1136/ precancerous conditions. Here we present a deep- were unaltered white-light images anonymised and gutjnl- 2019- 319347). learning (DL) approach for the diagnosis of atrophic exported as Digital Imaging and Communications Chair for Clinical gastritis developed and trained using real- world in Medicine (DICOMs). Non- standardised images Bioinformatics, Saarland endoscopic images from the proximal stomach. The (eg, various scope positions, distances, angles and University, Saarbrücken, Germany model achieved an accuracy of 93% (area under illumination; bile, food and mucus contaminations) Department of Medicine II, the curve (AUC): 0.98; F- score 0.93) in an inde- were taken from the non- overinflated proximal Saarland University Medical pendent data set, outperforming expert endosco- stomach (gastric corpus and fundus). All images Center, Saarland University, pists. DL may overcome conventional 
TI - Deep-learning based detection of gastric precancerous conditions
JF - Gut
DO - 10.1136/gutjnl-2019-319347
DA - 2020-01-02
UR - https://www.deepdyve.com/lp/british-medical-journal/deep-learning-based-detection-of-gastric-precancerous-conditions-v0VBNGCxV2
SP - 4
EP - 6
VL - 69
IS - 1
DP - DeepDyve
ER -