The diagnosis of invasion depth of gastric cancer estimates how deeply the cancer invades gastric mucosa. Although it plays an indispensable role to the choice of treatment methods, it is known to be difficult even for expert endoscopists. On the other hand, the recent development of deep neural networks (DNNs) has achieved image recognition ability outperforming humans. Therefore, it is the purpose of our research to promote its accuracy, applying the intelligent ability of DNNs to the diagnosis of the invasion depth of gastric cancer based on endoscopic images. Cooperating with the doctors of the Okayama University Hospital, we have achieved the accuracy over 80% until now. Also, our research is in progress to further improve the accuracy for practical use.

Fig. An example of the distribution of the features of gastric cancer in their 1st and 2nd principal component space