Fig. 3
From: Deep learning for gastroscopic images: computer-aided techniques for clinicians
(Reproduced with permission from Ref. [37]. Copyright 2020 Springer Nature Publishing)
Diagnosis with a CAD system for the endoscopic video of a post-eradication subject. The computer-aided diagnosis system for white-light imaging (WLI-CAD, upper side) returned a prediction value of 0.492 for a post-eradication status, which turned out to be an incorrect prediction. However, linked colour imaging (LCI-CAD, lower side) returned a prediction value of 0.985 for a post-eradication status, which turned out to be the correct prediction. The lower heatmap demonstrates that hot spots were drawn against the contrast between a pale reddish tone and a whitish tone of the gastric mucosa in the captured LCI image.