Fig. 12From: Deep learning for gastroscopic images: computer-aided techniques for clinicians(Reproduced with permission from Ref. [2]. Copyright 2021 Elsevier Publishing)EGD images classified into 31 sites and representative images identified by IDEA. The images show 31 sites determined by IDEA and the prediction confidence. Class 0, hypopharynx; 1, oesophagus; 2, gastroesophageal junction; 3, gastric cardia P in the antegrade view; 4–7, lower body (A, G, P, L); 8–11, middle-upper body in the antegrade view (A, G, P, L); 12–15 gastric cardia in the retroflex view (A, G, P, L); 16–18, middle-upper body in retroflex view (A, P, L); 19–21, angularis (A, P, L); 22–26, antrum (whole, A, G, P, L); 27, pylorus; 28–29, duodenal bulb (A, P); 30, duodenal descending. A, anterior wall; G, greater curvature; P, posterior wall; L, lesser curvature.Back to article page